Medical Diagnostics AI
- David Mays
- Dec 12, 2023
- 28 min read
Updated: Jan 1, 2024
Article Sections
(1) Medical Diagnostics
Medical Diagnostics is the diagnosis or clarification of exact conditions or illness that cause the symptoms that patients present with when seeking medical healthcare. After Diagnostics has determined the classification of ailment a patient possesses, diagnosticians will then identify both the type and progression of said condition if variances of type and progression apply. Through the course of this examination, the premium appropriate effectual course of treatment applicable to the particular variance of the patients ailment should become apparent. With some classifications of illness the patients own individual genetic makeup can also factor in deciding capital course of treatment.
Diagnostic testing is then also subsequently used to determine the effectiveness of the selected treatment regarding removal of the illness from which the patient is suffering. Within Diagnostics the foremost methods of testing used in patient analysis are split into 3 different fields:
(1) Clinical Pathology
(2) Anatomical Pathology
(3) Radiological testing
This article contains a review and description of the majority of the main methods of testing and diagnostics processes within Clinical Pathology and Anatomical Pathology.
For the purpose of contributing to current and future AI research in the field of Medical Diagnostics, we have conducted a review and filtering of the best and updated AI research papers within Medical Diagnostics that we could locate. The AI research papers have been separated according to their relevant fields within Clinical Pathology and Anatomical Pathology and are listed at the end of each diagnostics subsection. Radiological testing and relevant Radiological AI research will be added to this article in the near future.
If you are not familiar with Artificial Intelligence and its relevance to being applied to other fields of research you can read our article on The Definition Of AI here.
(2) Medical Diagnostics AI
What has AI achieved in the context of medical diagnostics?
The AI machine learning (ML) model automation of medical diagnostic interpretation of patient symptoms and causal condition has reached the point of matching if not surpassing human physician rates of success in positively diagnosing patient illness.
In the 2019 study "A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis", Denniston et al conducted a review of 25 separate radiological examination studies that compared the diagnostic performance of AI ML diagnostic models against the diagnostic performance of human physicians. The 25 radiological studies selected to perform the comparative review primarily focused on singular illness disciplines including cancer, ophthalmology, neurology, respiratory disorders, trauma and orthopedics, cardiology, dermatology, gastroenterology. Some of the studies reviewed contained datasets featuring tens of thousands of radiological images that were used to train the AI ML diagnostic models. Denniston et al utlised specificity and sensitivity scores to rate the accuracy and therefore differentiate the performance of the AI ML models versus the performance of human physicians. Sensitivity equates to the probability of a diagnostic test successfully diagnosing patients who posses the tests target illness (0-100%), while specificity measures the accuracy of the diagnostic test in successfully identifying the patients who do not posses the tests target illness (0-100%). The final conclusion of Denniston et al was that out of the 25 radiological studies reviewed, the AI ML models achieved overall sensitivity of 87% and specificity of 93%, while the human physicians achieved overall sensitivity of 86% and specificity of 91%.
In the 2019 study "Artificial Intelligence Versus Clinicians in Disease Diagnosis: Systematic Review", Ming et al conducted a comparative review of 9 different radiological AI diagnostic research papers to examine the differences in medical diagnostic performance between AI ML models and human physicians. The 9 medical AI studies selected to perform the comparative review primarily focused on singular illness disciplines including ophthalmology, dermatology, cancer, thoracic disorders, neurological disorders. The datasets utilised to train the AI ML diagnostic models in the 9 different studies varied in quantity of radiological images, with a range from 211 images to over 100,000 images. The performance metrics utilised in the 9 different studies including accuracy, sensitivity, specificity, and false-positive rate. Ming et al reported that in all 9 studies reviewed, the performance of the AI ML diagnostic models largely matched or even surpassed the diagnostic performance of the human physicians in the same studies.
In the 2021 study "Artificial intelligence in disease diagnostics: A critical review and classification on the current state of research guiding future direction", Stieglitz et al conducted an extensive review of 126 separate AI medical studies carried out to determine the diagnostic capabilities of AI ML models in successfully diagnosing different illnesses. The datasets utilised for the AI ML models in the 126 different studies majoritively consisted of patient data from preexisting medical databases concerning cardiovascular disorders, neurological/psychiatric, gastrointestinal disease, infectious disorders. The largest dataset from the medical databases utilised for AI ML model training consisted of 212,554 patients. Stieglitz et al posited that the accuracy, specificity and sensitivity scores (the main metrics employed for determining positive diagnostic success rate) of the AI ML models utilised for all 126 studies were in between 60% to 100%.
In the 2023 study "Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda" Kumar et al conducted an extensive review of 54 separate AI medical studies carried out to determine the diagnostic capabilities of AI ML models in successfully diagnosing different illnesses. The datasets utilised for the AI ML models in the 54 different studies consisted of independent disease specific patient datasets or institutional disease databases concerning skin disease, liver disease, diabetes, urology disease, kidney disease, cancer, heart disease, cardiovascular disorders, Retinopathy disease, neurological/psychiatric disorders, gastrointestinal disease, infectious disorders. Kumar et al declared that the accuracy, specificity and sensitivity scores of the AI ML models utilised for all 54 studies were, on average, over 80%.
What are the challenges to developing AI deployment in medical diagnostics?
Looking forward, this breakthrough in automated machine learning diagnosis success rate remains inconsistent due to dataset variables and AI ML model variables that affect its ability to universally match or surpass human physician level diagnostic capabilities. The dataset variables and AI model variables include (a) the age, gender, sex, ethnicity of patients within datasets, (b) different amounts of patient data available for ML model research depending upon illness type, (c) different hospital diagnostic machinery, (d) different configurations of AI machine learning models and model algorithms, and finally (e) different performance metrics for analysing model efficacy. The dataset variable issues listed above, have the potential to decrease AI ML model diagnostic accuracy due to either the AI models not being trained with enough patient data from each of the data variable categories or due to the target patient scope of the AI models being too convex.
However coordinated efforts are being organised and exerted within the medical and AI industries in order to reduce the effect of those variables and therefore achieve maximum efficiency in medical machine learning diagnostics. These coordinated efforts are being organised and exerted at two levels; (1) the data organisation level concerning the availability and pooling of patient data between researching organisations, and at (2) the AI ML research stage concerning the design, coding and pooling of new or updated ML models that are created to replace specific singular medical diagnostic tests.
(1) Efforts to synchronise and globalise healthcare data for the purpose of AI research include the adoption and implementation of interoperability standards for Electronic Health Records (EHR). Electronic Medical Records (EMR) are the patient health records that are stored by individual health organisations such as hospitals, while Electronic Health Records (EHR) are the same patient health records but facilitate the sharing and pooling of the patient data records between different health organisations. Electronic interoperability standards have been produced by international standardization groups such as Health Level 7 (HL7) in order to promote, increase, simplify and improve the "exchange, integration, sharing and retrieval" of EHR data for the purpose of clinical healthcare and medical research. The Fast Healthcare Interoperability Resources (FHIR) are the electronic interoperability standards created by Health Level 7 (HL7).
Currently AI diagnostic researchers can obtain external healthcare data through 2 main data sources which are patient databases or publicly accessible datasets. Patient Databases consist of real world datasets or clinical trial datasets. Real world datasets consist of patient medical data produced by regular medical care carried out by medical clinics, hospitals and pharmacies. In order to utilise real world datasets for AI diagnostic research, the datasets need significant filtering and processing as the patient data consists of structured and unstructured medical records. Clinical trial datasets consists of data created by medical organisations when conducting controlled clinical trials within a specific medical field on potentially defined cross sections of patients. Clinical trial datasets are rarely made accessible to outside parties due to competitive and ethical self interests of the medical groups conducting the trials. Publicly accessible datasets on the other hand are published by medical research groups and hospitals or are published by public medical archives for example The Cancer Imaging Archive or The Influenza Research Database.
The data sources available to AI diagnostic researchers are also categorised into 3 further different classifications including data warehouses, data lakes and data silos. Data warehouses consist of accessible collections of healthcare data from multiple sources such as different hospital databases, EHR, EMR, radiology databases, laboratory databases, pharmacy records. Data warehouses process, store and secure structured healthcare data in a global format that allows AI researchers to access the stored data without the need for the AI researchers to conduct further processing on retrieved data. Data lakes are similar to data warehouses in that they process, store and secure accessible healthcare data from multiple sources but the data is stored in unstructured or half structured formats. Data silos are databases of healthcare data that usually belong to singular medical organisations (for example a hospitals internal database of patient records) that share the stored medical data within their own organisation but refrain from sharing the stored medical data with parties external to their organisation.
(2) Efforts to synchronise and globalise the design, coding and pooling of new or updated ML models include ML researchers adopting best practice standards including STARD AI (Standards for Reporting Diagnostic Accuracy), TRIPOD AI (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis), Claim (Checklist for Artificial Intelligence in Medical Imaging), Future-AI (Fairness, Universality, Traceability, Usability, Robustness and Explainability), Consort AI (Consolidated Standards of Reporting Trials). These standards effectively ensure that the studies and research conducted by AI medical diagnostic researchers remains transparent and complete, therefore allowing other AI researchers and medical physicians to study and evaluate the guided research and even medically deploy the guided research with confidence. The above standards inform AI researchers of best practice within vital research report section areas including title, abstract, methods, results, discussion, ground truth labels, data partitions, model description, training and evaluation, risk of bias, study design, data & optimization, model performance, model examination, model reproducibility.
References
A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis
https://www.sciencedirect.com/science/article/pii/S2589750019301232?via%3Dihub#!
Artificial Intelligence Versus Clinicians in Disease Diagnosis: Systematic Review
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6716335/
Artificial intelligence in disease diagnostics: A critical review and classification on the current state of research guiding future direction
https://d-nb.info/1243730676/34
Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8754556/
Addressing bias in big data and AI for health care: A call for open science
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8515002/
Artificial Intelligence in Healthcare: Perception and Reality
Canadian Association of Radiologists White Paper on Artificial Intelligence in Radiology
https://journals.sagepub.com/doi/10.1016/j.carj.2018.02.002
Key challenges for delivering clinical impact with artificial intelligence
https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-019-1426-2
A framework for validating AI in precision medicine: considerations from the European ITFoC consortium
https://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/s12911-021-01634-3
Reporting quality of studies using machine learning models for medical diagnosis: a systematic review
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7103817/
Guidelines for Reporting AI Research
https://pubs.rsna.org/page/ai/blog/2022/09/ryai_editorsblog0928
Protocol for development of a reporting guideline (TRIPOD-AI) and risk of bias tool (PROBAST-AI) for diagnostic and prognostic prediction model studies based on artificial intelligence
https://bmjopen.bmj.com/content/11/7/e048008
Developing a reporting guideline for artificial intelligence-centred diagnostic test accuracy studies: the STARD-AI protocol
https://bmjopen.bmj.com/content/11/6/e047709
Reporting of Artificial Intelligence Diagnostic Accuracy Studies in Pathology Abstracts: Compliance with STARD for Abstracts Guidelines
https://www.sciencedirect.com/science/article/pii/S2153353922000918
What are the solutions to the those challenges?
(Daylight Magazine Opinion)
(1) There needs to be an international database of diagnostic AI ML models for every disease specific diagnostic test available to pathologists and physicians. This database should contain the best performing efficient AI ML model configurations, corresponding dataset details, and corresponding performance metrics.
(2) The international database needs the machine learning models and corresponding patient datasets to then be subdivided further according to patient type (age, gender, sex, race) and medical equipment utilised (if medical equipment variables in singular datasets decrease AI model diagnostic performance).
What does the future of AI in medical diagnostics and medical treatment look like?
(Daylight Magazine Opinion)
Looking to the future we will depend on the continued progression of Moore's law (the progressive increase of transistor density relative to electronic chip size) so that we can reduce the component size and therefore price of the diagnostic machinery (for example a Coulter Counter or Clinical Chemistry Analyser) responsible for the biomedical or microscopic analysis of patient fluid and tissue samples and therefore create replacement consumer level diagnostic machinery. Consumers would then utilise their diagnostic machinery in conjunction with consumer AI machine learning (ML) model software that together will analyse and diagnose any fluid or tissue samples that they enter into their own diagnostic machinery. The future inexpensive consumer diagnostic machinery and AI ML software could be used to replace visits to community physicians or hospitals in the case of illness or suspected illness and therefore replace 2 week or even month long waiting for the results of hospital or laboratory diagnostics tests. There will also be a future availability of secondary treatment analysis AI ML software into which consumers can enter the diagnostic results from their own diagnostic system. The treatment analysis AI ML software would then suggest the best course of action/treatment (for example further tests, pharmaceutical medicine, surgery) that consumers should take in order to remove the illness that their consumer diagnostic system has identified. For the foreseeable future due to procedural safety reasons radiological diagnostics or internal patient tissue diagnostics would remain physician or surgeon orientated hospital procedures.
(3) Medical Diagnostic Fields and Relevant AI Research
Clinical Pathology
Clinical Pathology consists of 4 subdivisions:
- Hematology/Hematopathology
- Clinical Chemistry
- Immunology
- Medical Microbiology
Hematology
Hematopathology
Human blood is created in the Bone Marrow, Thymus and Spleen and then pumped around the body by the human heart in order to transport essential oxygen and nutrients to tissue and organs but also remove waste products. It regulates water to acidity levels and helps the immune system to fight illness. Blood consists of Plasma, Platelets, Red Blood Cells and White Blood Cells. Red Blood Cells distribute oxygen within Hemoglobin to all bodily tissue with the oxygen produced by the respiratory system. White blood cells are deployed throughout the body to fight infection representing an important part of immune system defense. Plasma is the fluid component of the blood in which blood cells and platelets are transported around the body. Plasma also contains vitamins, hormones, proteins, minerals, fats and sugars. Platelets are used in order to prevent or slow bleeding through blood clotting. White Blood Cells, Red Blood Cells and Platelets are all produced in the Bone Marrow.
Hematology/Hematopathology combined are the microscopic and chemical examination of the blood and the blood forming organs in order to diagnose and further observe diseases of the blood. Hematology involves the study, diagnosis and treatment of diseases related to hemoglobin, platelets, blood cells, blood proteins, blood vessels and blood forming organs. Hematopathology is the study, diagnosis and treatment of diseases that relate specifically to blood cells and blood cell production by the Bone Marrow, Spleen and Thymus. Hematology/Hematopathology can be used to diagnose cardiovascular conditions, liver disease, Wilson disease, Anemia, inflammatory disease, infection, bleeding disorders, coagulation functionality, cancers of the blood (Leukemia, Lymphoma, Myeloma), other cancers.
Common Hematology testing include the following:
Blood Cell Counts are examinations of the quantities of Red Blood Cells, White Blood Cells, Hemoglobin, Platelets, and Hematocrit present in the blood. Blood count tests can also include Mean Corpuscular Volume tests and Reticulocyte count tests. Blood cell count testing methods most frequently involve utilising an automated Hematology analyser, while up to 25% of Blood cell count tests can require microscopy analysis of blood smear samples. Automated Hematology analysers utilise the Coulter principle which involves blood count analysis through the technique of resistive pulse analysis. Resistive pulse analysis involves dissolving blood samples into electrolyte solution then passing the electrolyte solution with blood cells through the apparatus of the automated analyser which records the changes in electrical resistance produce by the motion of the blood cells. The changes in electrical resistance can be processed and interpreted to produce a chemical and physical analysis of the blood samples cellular content. Blood Cell Counts are conducted to test patients for Anemia, cancers of the blood, inflammatory disease, pathogen infection.
Blood Enzyme Testing measures the quantity of Enzymes that are present in the blood. Enzymes are proteins that assist the bodys metabolism and the bodys chemical changes. Different types of enzymes include Alkaline Phosphatase, Lipase, Peptidases, Acid Phosphatase, Amylase, Transaminases. Enzyme Marker tests are the main testing method for blood enzymes and are used to test for cardiovascular conditions, organ function, liver disease, brain cancer, Wilson disease and other conditions. The main enzyme markers that are analysed in enzyme marker tests include CPK (creatine phosphokinase) isoenzymes CPK 1, CPK 2, CPK 3, the heart enzyme troponin, the liver enzymes alanine aminotransferase (ALT) and aspartate aminotransferase (AST). Enzyme marker tests are conducted by obtaining patient blood samples and then carrying out an immunochemical assay of the patient blood sample. Immunochemical assays involve utilising the immune system defense mechanism of animals and humans where antibodies (proteins) will attach themselves to foreign antigens in order to remove them from animal or human bodies.
The most commonly utilised Immunochemical assay is an Enzyme-linked immunosorbent assay (ELISA). For ELISA Immunoassay testing, Monoclonal and Polyclonal antibodies are created and designed in laboratories. Microtiter plates are then utilised to mix the patient blood sample with the monoclonal and polyclonal antibodies which identify and attach themselves to the targeted enzyme separating it from the patient blood sample. The level of the targeted enzyme present in the patient blood sample is measured through a colour analysis of the resulting combined antibody and enzyme solution. Other Immunochemical assay types involve inducing the same antibody to antigen reaction in patient blood or fluid but also involve employing a different method of measuring the result of that reaction for example Radioimmunoassays (RIA), Fluoroimmnoassays (FIA), Chemiluminescence immunoassays (CLIA). Electrophoresis is another method of testing enzyme levels in patient blood samples. Electrophoresis involves separating blood particles according to their electrical charge therefore isolating blood enzymes for patient analysis and diagnosis.
Bone Marrow Testing is used to examine the health of the fluid and tissue in patient bone marrow and the blood cells it produces. Bone marrow tissue produces red blood cells, white blood cells and platelets while Bone Marrow liquid contains stem cells and produces vitamins for blood cell production. There are 2 types of Bone Marrow tests including (1) Bone Marrow aspiration tests which extracts Bone Marrow fluid and cells and (2) Bone Marrow biopsy tests which extract bone and bone marrow tissue. The extracted patient samples are then examined under microscopes to perform illness diagnostics. The microscopic examinations of the patient bone marrow samples determines whether the bone marrow has cancer including leukemia, multiple myeloma, lymphoma, polycythemia vera, or secondary cancers. Bone marrow tests also diagnose abnormal chromosomes, anemia, aplastic anemia, thrombocytopenia, thrombocytosis leukopenia, leukocytosis. Bone marrow testing can also diagnose preemptive signs of illness from bone marrow blood cell production or can also determine patient illness and treatment progression.
Prothrombin Time Testing (PT) is used in combination with the internationalised normalised ratio calculation and partial thromboplastin time testing (PTT) to examine the time it takes for the plasma proteins known as "coagulating factors" in your blood to produce coagulation (clotting). The human body first employs platelets at the site of blood vessel tears as temporary clotting followed by plasma proteins which form fibrin in order to permanently repair the damaged blood vessel. The PT test and PTT test involve extracting a patient blood sample for laboratory analysis where the chemical process which causes blood clotting (hemostasis) is replicated in a test tube by adding chemical reagents to the patient blood sample. The time recorded for completing the coagulation process in the test tube blood sample demonstrates the condition of the patients coagulation functionality. Prothrombin Time and partial thromboplastin time tests can be used to detect bleeding disorders, measure warfarin (blood thinning medicine)dosage effect on patients, liver disease, vitamin k deficiency.
Common Hematopathology testing include the following:
Chromosome Banding Analysis is a cytogenetics method of testing a patients blood cell chromosomes for structural changes or other abnormalities that confirm diagnosis of specific illness. The 23 Chromosome structures consist of DNA and proteins which constitute the inherited genetic instructions to the human bodies formation and are contained within the nuclei of human cells. Chromosome Banding analysis involves staining the 23 chromosome structures with dye so that the chromosomes present in different colours which highlight the structural changes and abnormalities diagnosticians are looking for during microscopy analysis. Chromosome Banding Analysis can be used to diagnose genetic disorders, cancer, and birth defects.
Flow Cytometry is a laser based testing method that measures the physical and chemical properties of cells and particles. A Flow Cytometry machine examines patient samples of blood, disintegrated tissue, or cell cultures in order to determine cell characteristics, cell numbers, cell types, biomarker presence, microorganism presence, DNA gene expression. The patient sample cells are suspended in fluid and placed in a flow cytometer analyser which utilises hydrodynamic focusing to ideally pass the cells singularly through the laser in the flow cytometer. The cells are passed through the laser at a rate of 1000 cells per second and the patterns in which the passing cells disperses light from the laser are recorded and then digitally converted to allow pathologists to analyse the aforementioned cell characteristics for patient diagnostics. The results from this test will diagnose cellular health, cancer (Leukemia, Lymphoma, Myeloma), cytokine production, virus infected cells, cardiac disease and other diseases.
Immunohistochemistry in hematopathology is the use of fluorescent staining antibodies that attach themselves to specific antigen molecules (proteins) in blood cells in order to determine microscopic confirmation of the specific antigen molecules presence. Antibodies (Polyclonal or Monoclonal) are produced naturally by the human bodys immune system in order to locate, attach to and remove any foreign antigens (bacteria, viruses, parasites, fungi) or organic antigens (cancer) from the human body; therefore the introduction of laboratory produced antibodies to a patient sample can aid pathologists in locating antigens within the patient sample resulting in positive diagnosis of patient illness. Pathologists utilise different antibody staining techniques (Chromogenic immunohistochemistry, or Immunofluorescence) to ensure that the antibodies attach to and demonstrate clearly any antigen location within the patient sample during microscopic observation. Immunohistochemistry is utilised in the diagnosis of various cancers including lymphoma, prostate, colorectal, lung, bladder, pancreatic, and ovarian. Pathologists also employ Immunohistochemistry to diagnose hematologic malignancy, infectious diseases, neurodegenerative disorders and conduct proteome mapping.
Hematology Hematopathology AI Research
Applied machine learning in hematopathology
Artificial intelligence and its applications in digital hematopathology
Artificial intelligence in hematological diagnostics: Game changer or gadget?
https://www.sciencedirect.com/science/article/pii/S0268960X22000935?via%3Dihub#s0025
Machine learning approach of automatic identification and counting of blood cells
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6718065/
Challenging gold standard hematology diagnostics through the introduction of whole genome sequencing and artificial intelligence
https://onlinelibrary.wiley.com/doi/full/10.1111/ijlh.14033
Scalable Prediction of Acute Myeloid Leukemia Using High-Dimensional Machine Learning and Blood Transcriptomics
https://www.sciencedirect.com/science/article/pii/S2589004219305255?pes=vor#abs0020
Artificial intelligence to assist specialists in the detection of haematological diseases
https://www.cell.com/heliyon/pdf/S2405-8440(23)03147-X.pdf
Machine Learning Approach to Forecast Chemotherapy-Induced Haematological Toxicities in Patients with Rhabdomyosarcoma
https://www.mdpi.com/2072-6694/12/7/1944
Machine learning approaches classify clinical malaria outcomes based on haematological parameters
https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-020-01823-3
An application of machine learning to haematological diagnosis
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5765139/
A Review of Artificial Intelligence Applications in Hematology Management: Current Practices and Future Prospects
https://www.jmir.org/2022/7/e36490/
Predicting Complete Remission of Acute Myeloid Leukemia: Machine Learning Applied to Gene Expression
https://journals.sagepub.com/doi/epub/10.1177/1176935119835544?src=getftr
Clinically validated machine learning algorithm for detecting residual diseases with multicolor flow cytometry analysis in acute myeloid leukemia and myelodysplastic syndrome
https://www.sciencedirect.com/science/article/pii/S2352396418304602?pes=vor#s0025
Machine learning can identify newly diagnosed patients with CLL at high risk of infection
https://www.nature.com/articles/s41467-019-14225-8#Sec15
Machine Learning Applications in the Diagnosis of Benign and Malignant Hematological Diseases
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8432325/
Artificial Intelligence in Haematology
https://jhscr.org/index.php/JHSCR/article/view/78
Integrating artificial intelligence into haematology training and practice: Opportunities, threats and proposed solutions
https://onlinelibrary.wiley.com/doi/full/10.1111/bjh.18343
How artificial intelligence might disrupt diagnostics in hematology in the near future
https://www.nature.com/articles/s41388-021-01861-y
Machine learning and artificial intelligence in haematology
https://onlinelibrary.wiley.com/doi/full/10.1111/bjh.16915
A Review of Artificial Intelligence Applications in Hematology Management: Current Practices and Future Prospects
Clinical Chemistry
Clinical Chemistry is the use of biochemical analysis in order to determine the presence and nature of chemical compounds (known as analytes or biomarkers) in bodily fluids. The measurement of the presence and characteristics of these chemical compounds in humans can determine the presence of a particular illness and the progressive stage of that illness. This biochemical analysis of blood plasma, blood serum, urine, or cerebrospinal fluid involves two main methods of either (a) inducing basic chemical reactions in bodily fluids to determine routine diagnostic compounds or (b) employing advanced measurement techniques to detect compounds or detect the varying characteristics and quantities of these biomarkers.
The biomarkers that are analysed during these tests include electrolytes, hormones, carbohydrates, lipids, enzymes, proteins. The different forms and levels of these biomarkers present in patient blood, urine or cerebrospinal samples can demonstrate the particular illness a patient is presently undergoing. The advanced techniques are capable of examining normal properties of biomarkers but also in increased detail for example the biomarkers relation to light, or the biomarkers electronic voltage properties. These advanced techniques include immunoassays, electrophoresis, spectrometry, chromatography, spectrophotometry, potentiometry, fluorometry, nephelometry and turbidimetry.
Clinical Chemists advise clinical physicians as to what sample extraction and tests need to be employed to diagnose, monitor or prevent patient illness and will then advise physicians regarding the interpretation of test results. Clinical Chemists may also have to process patient samples before the tests are conducted depending upon the method being utilised. The Clinical Chemistry laboratories now include automated analysers which conduct a range of tasks from basic specimen analysis to the advanced measurement techniques. Clinical Chemistry can be used to assist in the diagnosis of Heart disease, Liver disease, Pancreas disorders, Diabetes and other Endocrinological disorders, Kidney disease, Autoimmune disease, Blood cancer, Gland functionality.
Clinical Chemistry AI Research
Machine Learning in Clinical Pathology:Seeing The Forest For The Trees
https://academic.oup.com/clinchem/article/64/11/1553/5608645
Using Machine Learning to Aid the Interpretation of Urine Steroid Profiles
https://academic.oup.com/clinchem/article/64/11/1586/5608626
How is test laboratory data used and characterised by machine learning models? A systematic review of diagnostic and prognostic models developed for COVID-19 patients using only laboratory data
https://www.degruyter.com/document/doi/10.1515/cclm-2022-0182/html#j_cclm-2022-0182_fig_003
Artificial Intelligence and Mapping a New Direction in Laboratory Medicine: A Review
https://academic.oup.com/clinchem/article/67/11/1466/6374790
Using Machine Learning to Predict Laboratory Test Results
https://academic.oup.com/ajcp/article/145/6/778/2836697
Has the Flood Entered the Basement? A Systematic Literature Review about Machine Learning in Laboratory Medicine
https://www.mdpi.com/2075-4418/11/2/372
Artificial Intelligence in Clinical Chemistry: Dawn of a New Era?
https://link.springer.com/article/10.1007/s12291-023-01150-3
Artificial intelligence: is it the right time for clinical laboratories?
https://www.degruyter.com/document/doi/10.1515/cclm-2022-1015/html?lang=en
Machine Learning For Clinical Chemists
https://academic.oup.com/clinchem/article/65/11/1350/5715874
Clinical chemistry in higher dimensions: Machine-learning and enhanced prediction from routine clinical chemistry data
https://www.sciencedirect.com/science/article/abs/pii/S0009912016301709
Using machine learning to develop an autoverification system in a clinical biochemistry laboratory
https://pubmed.ncbi.nlm.nih.gov/33554565/
Artificial Intelligence Applications in Clinical Chemistry
https://www.sciencedirect.com/science/article/abs/pii/S0272271222000634?via%3Dihub
Machine Learning in Laboratory Medicine: Recommendations of the IFCC Working Group
https://academic.oup.com/clinchem/article-abstract/69/7/690/7186579?redirectedFrom=fulltext
How Can We Ensure Reproducibility and CLinical Translation of Machine Learning Applications in Laboratory Medicine?
https://academic.oup.com/clinchem/article/68/3/392/6501535
A Machine Learning Approach for the Automated Interpretation of Plasma Amino Acid Profiles
https://academic.oup.com/clinchem/article/66/9/1210/5900235
Automation and artificial intelligence in the clinical laboratory
https://www.tandfonline.com/doi/full/10.1080/10408363.2018.1561640
The multicenter European Biological Variation Study (EuBIVAS): a new glance provided by the Principal Component Analysis (PCA), a machine learning unsupervised algorithms, based on the basic metabolic panel linked measurands
https://www.degruyter.com/document/doi/10.1515/cclm-2021-0599/html
Machine Learning for the Biochemical Genetics Laboratory
https://academic.oup.com/clinchem/article/66/9/1134/5891797
Turnaround time prediction for clinical chemistry samples using machine learning
https://www.degruyter.com/document/doi/10.1515/cclm-2022-0668/html
Applications of machine learning in the chemical pathology laboratory
https://jcp.bmj.com/content/74/7/435
Immunology
The Immune System is the human bodys defense system against any pathogenic organisms or foreign substances that cause illness within the human body. The Immune System primarily comprises of the Lymph System which is a collection of organs, tissue and cells. The Lymph organs which include the Thymus, Bone Marrow and Spleen produce White Blood Cells which the Immune System uses to combat illness. White blood cells are transported within the lymph fluid in order to be entered into the blood stream. Lymph fluid is produced by the Lymph glands and distributed around the body by the Lymph vessels similar in structure to the blood vessels.
Immunology is the study of the bodys immune system and the immune systems ability to differentiate and nullify foreign antigens or pathogens that threaten healthy bodily functionality. Immunological disorders affect the functionality of immune systems in their ability to prevent and remove illness but also their ability to self-regulate which can lead to immune systems attacking their own host body. There are 4 main different types of immune system disorders; Primary immunological deficiencies, Secondary immunological deficiencies, Hypersensitivities, Autoimmune disorders.
Primary Immune Deficiencies (PID) are genetically inherited conditions that can inhibit both the innate and adaptive immune systems first presenting during childhood. PIDs inhibit white blood cell operation making your body more vulnerable to infection and disease. Primary Immune Deficiencies (PIDs) include Severe Combined Immunodeficiency Disease (SCID), Wiskott-Aldrich syndrome (WAS), Common Variable Immunodeficiency (CVID), Ataxia telangiectasia. Methods of testing for PIDs include Blood Cell counts, Lymphocyte tests, Granulocyte tests, B cell tests, Immunoglobulin measurements.
Secondary Immune Deficiencies (SID) are immune disorders that develop later in life due to secondary causes including blood or bone marrow disorders, medicinal drugs, malnutrition, bacterial or viral infections. SIDs also inhibit the white blood cells and pathways in their removal of illness. SIDs include Antibody deficiencies, Neutropenia, T-cell deficiency. Methods of testing for SIDs include measuring T cells, B cells, monocytes, neutrophils, natural killer cells, antibody responses and immunoglobulins through lymphocyte phenotyping and blood testing.
Hypersensitivity Disorders consist of the immune system overreacting to otherwise benign external substances which can lead to the damage of bodily tissue in what medicine defines as allergic response. Hypersensitivity disorders include Asthma for which testing is Immunoglobulin testing, Food Allergies for which testing is dermatological excitation, Autoimmune Hemolytic anemia for which testing is a Coombs test (Immunoglobulins and C3 blood test).
Autoimmune Disorders are where an immune system is unable to discern between the bodys cells and foreign antigens which can lead to that immune system attacking its own host body. Autoimmune disorders include Multiple sclerosis, Rheumatoid arthritis, Type I diabetes, Lupus. Methods of testing for Autoimmune disease include C reactive protein testing, Erythrocyte Sedimentation Rate, Ferritin testing, Rheumatoid factor testing.
AI Research of Immunological Profiling of Diseases and Immune system responses
Deep learning of immune cell differentiation
https://www.pnas.org/doi/10.1073/pnas.2011795117
Artificial neural networks for immunological recognition
https://arxiv.org/ftp/arxiv/papers/1808/1808.03386.pdf
Applications of Machine and Deep Learning in Adaptive Immunity
https://www.annualreviews.org/doi/full/10.1146/annurev-chembioeng-101420-125021
Dana-Farber repository for machine learning in immunology
https://www.sciencedirect.com/science/article/abs/pii/S0022175911001773
Weighted gene coexpression network analysis and machine learning reveal oncogenome associated microbiome plays an important role in tumor immunity and prognosis in pan-cancer
https://translational-medicine.biomedcentral.com/articles/10.1186/s12967-023-04411-0
Machine learning methods and harmonized datasets improve immunogenic neoantigen prediction
https://www.cell.com/immunity/pdf/S1074-7613(23)00406-5.pdf
Machine-Learning-Assisted Analysis of TCR Profiling Data Unveils Cross-Reactivity between SARS-CoV-2 and a Wide Spectrum of Pathogens and Other Diseases
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9598299/
Integration of mechanistic immunological knowledge into a machine learning pipeline improves predictions
https://www.nature.com/articles/s42256-020-00232-8
Machine Learning Methods for Cancer Immunology
https://www.repository.cam.ac.uk/items/a154ee90-c421-4398-be6f-31b7ccf3c87c
Can we predict T cell specificity with digital biology and machine learning?
https://www.nature.com/articles/s41577-023-00835-3#Bib1
A Study of the Recent Trends of Immunology: Key Challenges, Domains, Applications, Datasets, and Future Directions
https://www.mdpi.com/1424-8220/21/23/7786
Combining immunoprofiling with machine learning to assess the effects of adjuvant formulation on human vaccine-induced immunity
https://www.tandfonline.com/doi/full/10.1080/21645515.2019.1654807
The immune system, adaptation, and machine learning
https://www.sciencedirect.com/science/article/abs/pii/016727898690240X
Computational modelling of immunological mechanisms: From statistical approaches to interpretable machine learning
https://www.sciencedirect.com/science/article/pii/S2667119023000095
Disease diagnostics using machine learning of immune receptors
https://www.biorxiv.org/content/10.1101/2022.04.26.489314v3.full
The Immune System Computes the State of the Body: Crowd Wisdom, Machine Learning, and Immune Cell Reference Repertoires Help Manage Inflammation
https://www.frontiersin.org/articles/10.3389/fimmu.2019.00010/full
Unsupervised machine learning reveals key immune cell subsets in COVID-19, rhinovirus infection, and cancer therapy
https://elifesciences.org/articles/64653
New tools for MHC research from machine learning and predictive algorithms to the tumour immunopeptidome
https://onlinelibrary.wiley.com/doi/full/10.1111/imm.12956
Comprehensive Data Integration Approach to Assess Immune Responses and Correlates of RTS,S/AS01-Mediated Protection From Malaria Infection in Controlled Human Malaria Infection Trials
https://www.frontiersin.org/articles/10.3389/fdata.2021.672460/full
Opportunities and Challenges in Democratizing Immunology Datasets
https://www.frontiersin.org/articles/10.3389/fimmu.2021.647536/full
Immunological Disorders AI Research
Artificial Intelligence and the Hunt for Immunological Disorders
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7908683/
A systematic review of the applications of artificial intelligence and machine learning in autoimmune diseases
https://www.nature.com/articles/s41746-020-0229-3#Sec13
Rapid, High-Throughput Single-Cell Multiplex In Situ Tagging (MIST) Analysis of Immunological Disease with Machine Learning
https://pubs.acs.org/doi/abs/10.1021/acs.analchem.3c01157
Early Diagnosis of Primary Immunodeficiency Disease Using Clinical Data and Machine Learning
https://www.sciencedirect.com/science/article/pii/S2213219822009254
Machine Learning Approaches in Study of Multiple Sclerosis Disease Through Magnetic Resonance Images
https://www.frontiersin.org/articles/10.3389/fimmu.2021.700582/full
Overview of microbial therapeutics in immunological disorders
https://www.sciencedirect.com/science/article/abs/pii/B9780323993364000082
Kinase Inhibitors for the Treatment of Immunological Disorders: Recent Advances
https://pubs.acs.org/doi/abs/10.1021/acs.jmedchem.8b00667
Seeing the Forest for the Trees: Evaluating Population Data in Allergy-Immunology
https://www.sciencedirect.com/science/article/abs/pii/S2213219821010102
Identification of Diagnostic Signatures and Immune Cell Infiltration Characteristics in Rheumatoid Arthritis by Integrating Bioinformatic Analysis and Machine-Learning Strategies
https://www.frontiersin.org/articles/10.3389/fimmu.2021.724934/full
Machine Learning Techniques for Personalised Medicine Approaches in Immune-Mediated Chronic Inflammatory Diseases: Applications and Challenges
https://www.frontiersin.org/articles/10.3389/fphar.2021.720694/full
A machine learning approach on whole blood immunomarkers to identify an inflammation-associated psychosis onset subgroup
https://www.nature.com/articles/s41380-022-01911-1
Artificial intelligence and deep learning to map immune cell types in inflamed human tissue
https://www.sciencedirect.com/science/article/abs/pii/S0022175922000205
Accurate Machine Learning Model to Diagnose Chronic Autoimmune Diseases Utilizing Information From B Cells and Monocytes
https://www.frontiersin.org/articles/10.3389/fimmu.2022.870531/full
Predictive Metagenomic Analysis of Autoimmune Disease Identifies Robust Autoimmunity and Disease Specific Microbial Signatures
https://www.frontiersin.org/articles/10.3389/fmicb.2021.621310/full
A machine learning approach to predict response to immunotherapy in type 1 diabetes
https://www.nature.com/articles/s41423-020-00594-4
Artificial Intelligence in Allergy and Immunology: Comparing Risk Prediction Models to Help Screen Inborn Errors of Immunity
https://karger.com/iaa/article/183/11/1226/824165/Artificial-Intelligence-in-Allergy-and-Immunology
Finding Gene Regulatory Networks in Psoriasis: Application of a Tree-Based Machine Learning Approach
https://www.frontiersin.org/articles/10.3389/fimmu.2022.921408/full
A Framework for Augmented Intelligence in Allergy and Immunology Practice and Research—A Work Group Report of the AAAAI Health Informatics, Technology, and Education Committee
https://www.sciencedirect.com/science/article/pii/S221321982200143X
A validated artificial intelligence-based pipeline for population-wide primary immunodeficiency screening
https://www.sciencedirect.com/science/article/abs/pii/S0091674922013434
Machine Learning-Based Gene Prioritization Identifies Novel Candidate Risk Genes for Inflammatory Bowel Disease
https://academic.oup.com/ibdjournal/article/23/9/1516/4560715
Medical Microbiology
Medical Microbiology involves the study of infectious diseases caused by exogenous (external origin) or endogenous (internal origin) pathogens that are also known as microorganisms. These microorganisms or pathogens include bacteria, fungi, viruses, parasites and prions that cause infectious diseases after invasion of human biology. Medical Microbiology conducts examination of these microorganisms and resulting infectious diseases in order to produce disease prevention, diagnosis and treatment capability.
The points at which these pathogens can enter the human biological system and cause infectious disease include the skin, external orifices, mucous membranes, and genitalia. The microorganisms can access these points of entry through (1) contact with infected humans or their infectious discharge, (2) contact with surfaces/substances or presence in environments that have been compromised with infectious microorganisms. The different microorganism types (bacteria, fungi, viruses, parasites) differ in the varying infectious diseases they can create, how those diseases present themselves, how the diseases progress, how the diseases are tested for and how those diseases are treated.
The examination and diagnosis of humans afflicted by suspected infectious disease is carried out through (1)microbiological cultures, (2) microscopic examination, (3) biochemical tests, (4) molecular tests, and (5) radiological imaging to determine diagnosis.
(1) Microbiological cultures involve attempting to grow, identify and examine infectious disease microorganisms from patient fluid or blood samples using one of 3 different mediums including solid cultures, liquid cultures or cell cultures. (2) Microscopic examination is used in conjunction with microbiological cultures. Different types of microscopy used in culture examination include Electron microscopes which pass electrons through cultures to increase cell inspection capability, and Compound Light microscopes which allow rapid initial inspection of cultures with chemical staining methods.
(3) Biochemical tests include serological methods which involve testing patient blood for the presence of immune system antibody responses (specifically the antibody Immunoglobulin) to the antigens of bacteria, fungi, viruses or parasites. Biochemical tests also include testing for enzymes produced by bacteria grown in liquid or solid cultures, for example Catalase tests or Oxidase tests. (4) The Molecular test most employed to diagnose pathogen related infectious disease is the Polymerase Chain Reaction which multiplies small samples of DNA into potentially billions of copies to obtain a sufficient DNA amount for examination. PCR involves using cultures of patient fluid samples to identify the presence of the genetics of the aforementioned pathogens and therefore diagnose an infectious disease.
Infectious Diseases targeted by Medical Microbiology include (by order of pathogen);
Viruses; Common cold, Influenza, Hepatitis, Respiratory syncytial virus, Rotavirus, Stomach flu, Human papillomavirus, COVID-19.
Bacteria; Tuberculosis, Strep throat, Typhoid Fever, Whooping cough, Sexually transmitted infections, Urinary tract infections, Salmonella, E. coli.
Fungi; Ringworm, Histoplasmosis, Fungal foot/nail infections, Thrush, Candida, Dandruff.
Parasites; Giardiasis, Hookworms, Pinworms, Malaria, Toxoplasmosis.
Medical Microbiology AI Research
Machine Learning for Antimicrobial Resistance Prediction: Current Practice, Limitations, and Clinical Perspective
https://journals.asm.org/doi/full/10.1128/cmr.00179-21
Machine Learning Advances in Microbiology: A Review of Methods and Applications
https://www.frontiersin.org/articles/10.3389/fmicb.2022.925454/full
Machine Learning and Applications inn Microbiology
https://academic.oup.com/femsre/article/45/5/fuab015/6174022
Machine learning in the clinical microbiology laboratory: has the time come for routine practice?
https://www.sciencedirect.com/science/article/pii/S1198743X20300859
Machine learning for clinical decision support in infectious diseases: a narrative review of current applications
https://www.sciencedirect.com/science/article/pii/S1198743X1930494X
Artificial intelligence, machine learning and deep learning: Potential resources for the infection clinician
https://www.sciencedirect.com/science/article/pii/S0163445323003791#bibliog0005
Applications of Machine Learning in Human Microbiome Studies: A Review on Feature Selection, Biomarker Identification, Disease Prediction and Treatment
Digital microbiology
https://www.sciencedirect.com/science/article/pii/S1198743X20303670
Machine learning in infection management using routine electronic health records: tools, techniques, and reporting of future technologies
https://www.sciencedirect.com/science/article/pii/S1198743X20300823
Predicting Antibiotic Resistance in Hospitalised Patients by Applying Machine Learning to Electronic Medical Records
https://academic.oup.com/cid/article/72/11/e848/5929660
Evolution of Diagnostic and Forensic Microbiology in the Era of Artificial Intelligence
Machine learning for microbiologists
https://www.nature.com/articles/s41579-023-00984-1
Machine learning on the road to unlocking microbiota’s potential for boosting immune checkpoint therapy
https://www.sciencedirect.com/science/article/pii/S1438422122000133
Applications of Machine Learning to the Problem of Antimicrobial Resistance: an Emerging Model for Translational Research
https://journals.asm.org/doi/full/10.1128/jcm.01260-20
A Framework for Effective Application of Machine Learning to Microbiome-Based Classification Problems
https://journals.asm.org/doi/full/10.1128/mbio.00434-20
Antimicrobial resistance and machine learning: past, present, and future
https://www.frontiersin.org/articles/10.3389/fmicb.2023.1179312/full
Machine Learning and Deep Learning Based Computational Approaches in Automatic Microorganisms Image Recognition: Methodologies, Challenges, and Developments
https://link.springer.com/article/10.1007/s11831-021-09639-x
Mini Review: Clinical Routine Microbiology in the Era of Automation and Digital Health
https://www.frontiersin.org/articles/10.3389/fcimb.2020.582028/full
Applications of Machine Learning in Human Microbiome Studies: A Review on Feature Selection, Biomarker Identification, Disease Prediction and Treatment
Machine Learning Takes Laboratory Automation to the Next Level
https://journals.asm.org/doi/full/10.1128/jcm.00012-20
(4) Medical Diagnostic Fields and Relevant AI Research
Anatomical Pathology
Anatomical Pathology consists of 3 divisions:
- Surgical Pathology
- Cytopathology
- Molecular Pathology
The Anatomical Pathology process in Diagnostics is concerned with removing and observing samples of the patients anatomy. The first stage of Anatomical Pathology involves Surgical Pathology which obtains tissue samples from patients through biopsies or surgery. Surgical Pathology then progresses to examining the tissue samples in order to diagnose or monitor illness. The different methods of examining tissue samples within Surgical Pathology include Gross Examination and Histopathological methods which include Histology Microscopy, Electron Microscopy, and Immunohistochemistry. The Anatomical Pathology divisions of Cytopathology and Molecular Pathology can also be used to observe patient tissue samples at increased microscopic depth. Cytopathology is the microscopic examination of patient tissue at the cellular level, while Molecular Pathology is microscopic tissue examination at the molecular level.
Surgical Pathology
- Surgical Specimen Extraction
/Tissue Biopsy
- Gross Examination
- Histopathology Preparation
- Histopathology Examination
To repeat, the first stage of Surgical Pathology and therefore Anatomical Pathology is the procurement of tissue samples from the anatomy of patients. These samples can be obtained through surgical procedures that produce surgical tissue specimens or non surgical biopsy methods that yield tissue extracts. Surgical methods involve removing tissue samples from organs or more substantially removing sections of organs during surgery. Biopsy methods mostly involve incisional biopsies which withdraw small samples from the affected area of the patients anatomy. The second stage of Surgical Pathology is Gross Examination. Gross Examination involves the pathologist studying the surgical specimen or tissue biopsy with their own eyes, unaided by microscopes. This examination can be productive with large tissue samples where visible illness symptoms can be observed and diagnosed by the pathologist. Gross Examination can also be used to select the areas of illness affected tissue that should be subject to histopathological examination.
The third stage of Surgical Pathology is preparing the tissue for Histopathological examination of the tissue. Histopathology is the field of medicine concerned with the diagnosis of illness through the microscopic study of diseased tissue. The tissue biopsy or surgical specimen will first be prepared and processed chemically in order to make diagnosis and observation of illness within the tissue more facile. Histopathological preparation first involves preserving the tissue sample either in a chemical fixative or through freezing, in order to retain the cellular structure of the tissue sample throughout the staining and examination process. The most commonly used chemical fixative is formalin which is diluted formaldehyde. The tissue sample is next installed or encapsulated in paraffin wax which is then placed onto glass slides. The pathologist will begin applying different histological staining techniques to the paraffin wax in order to intensify illumination between the different components within the tissue samples. Hematoxylin and Eosin are the standard staining chemicals used in this Histopathology staining process.
The fourth stage of Surgical Pathology is employing one or more of the Histopathological observation methods of Histology Microscopy, Electron Microscopy, or Immunohistochemistry. Histology Microscopy is the tissue level microscopic examination of the histochemically stained tissue samples in order to diagnose or monitor illness. The microscopic examination of tissue after histochemical staining is very useful for pathologists in differentiating between normal tissue and diseased tissue including identifying illness type and stage. Immunohistochemistry (as previously mentioned) involves the use of fluorescent histochemical staining antibodies that attach themselves to specific molecules of diseased tissue in order to determine microscopic confirmation of the specific molecules presence and therefore the diagnosis of disease. Electron Microscopy is the examination of tissue samples utilising an electron microscope which holds magnification capability 1 million times stronger than standard light microscopes. The electron microscope directs electrons at histochemically stained tissue samples which results in the production of black and white images of the tissue.
Histopathological Examination can be used to diagnose/monitor the following illnesses; Infectious diseases, Uterine fibroids, Cancer, Crohn's disease, Ulcerative colitis, early genetic changes leading to illness.
Histopathology AI Research
Deep Learning Approaches in Histopathology
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9654172/
Recent Advances of Deep Learning for Computational Histopathology: Principles and Applications
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8909166/
Machine learning in computational histopathology: Challenges and opportunities
https://onlinelibrary.wiley.com/doi/full/10.1002/gcc.23177
HEAL: an automated deep learning framework for cancer histopathology image analysis
https://academic.oup.com/bioinformatics/article/37/22/4291/6278294
Deep Learning of Histopathology Images at the Single Cell Level
https://www.frontiersin.org/articles/10.3389/frai.2021.754641/full
Machine Learning Methods for Histopathological Image Analysis
https://www.sciencedirect.com/science/article/pii/S2001037017300867#s0005
Machine Learning Methods for Histopathological Image Analysis: A Review
https://www.mdpi.com/2079-9292/10/5/562
Deep Learning Approaches in Histopathology
https://www.mdpi.com/2072-6694/14/21/5264
Machine Learning Methods for Computer-Aided Breast Cancer Diagnosis Using Histopathology: A Narrative Review
https://www.sciencedirect.com/science/article/abs/pii/S193986541930551X
Machine learning in computational histopathology: Challenges and opportunities
https://onlinelibrary.wiley.com/doi/full/10.1002/gcc.23177
Objective Diagnosis for Histopathological Images Based on Machine Learning Techniques: Classical Approaches and New Trends
https://www.mdpi.com/2227-7390/8/11/1863
Self supervised contrastive learning for digital histopathology
https://www.sciencedirect.com/science/article/pii/S2666827021000992
Deep Learning in Histopathology: A Review
https://wires.onlinelibrary.wiley.com/doi/epdf/10.1002/widm.1439
Selection of diagnosis with oncologic relevance information from histopathology free text reports: A machine learning approach
https://www.sciencedirect.com/science/article/abs/pii/S1386505622000284
Deep neural network models for computational histopathology: A survey
https://www.sciencedirect.com/science/article/abs/pii/S1361841520301778
State of machine and deep learning in histopathological applications in digestive diseases
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8160628/
Breast Cancer Histopathology Image Classification Using an Ensemble of Deep Learning Models
https://www.mdpi.com/1424-8220/20/16/4373
Conventional Machine Learning versus Deep Learning for Magnification Dependent Histopathological Breast Cancer Image Classification: A Comparative Study with Visual Explanation
https://www.mdpi.com/2075-4418/11/3/528
A method for utilizing automated machine learning for histopathological classification of testis based on Johnsen scores
https://www.nature.com/articles/s41598-021-89369-z
Strategies to develop radiomics and machine learning models for lung cancer stage and histology prediction using small data samples
https://www.sciencedirect.com/science/article/pii/S1120179721002982
Cytopathology
Cytopathology is the diagnosis and study of illness through the microscopic examination of anatomy at the cellular level. The examination of anatomical cells is performed by obtaining singular cells from tissue samples. After the cells have been collected from the patients anatomy they are placed on slides and subjected to histochemical staining before microscopic observation. Alternatively the extracted cells are placed in preservative liquid and subjected to further refinement processing (SurePath or Thinprep methods) before being placed on slides, histochemically stained and microscopically observed. The cells can be extracted using either Exfoliative Cytology or Intervention Cytology.
Exfoliative Cytology consists of obtaining patient cells either through spontaneous exfoliation or mechanical exfoliation. Spontaneous exfoliation involves retrieving patient cells after they have been spontaneously exfoliated or shedded from the patients body. Examples of cell extraction through Spontaneous exfoliation include conducting respiratory cytology tests to retrieve cells from patient phlegm or sputum, obtaining cells from patient urine, or acquiring cells from any irregular patient bodily discharge. Mechanical Exfoliation pertains to acquiring cells by physically scraping cells from the exterior of patient anatomy. Examples of mechanical exfoliation include brushing cells from the patients skin or mucous membranes (mouth, nose), conducting an endoscopy to scrape cells from the lining of the patients gastrointestinal tract (stomach and intestines), conducting a Gynecological Pap Smear which involves brushing cells from a patients cervix.
Intervention Cytology refers to extracting patient cells through employing methods of intervening into the patients anatomy. The normal intervening technique utilised in Intervention Cytology is fine needle aspiration. Fine needle aspiration consists of using a thin needle and syringe to pierce a patients anatomy to collect cells. Fine needle aspiration will be utilised to retrieve patient cells from any lesions, cysts or masses that are located under the skin or in patient bodily organs. Fine needle aspiration is also deployed to obtain cells from immune system lymph nodes, pleural fluid from the space between lungs and chest, or pericardial fluid in the cardiac sac. This intervention cytology technique is performed with the assistance of ultrasounds/CAT scans for intervening to targeted areas of internal bodily depth or with the assistance of palpitation guidance where the pathologist can manually feel the optimal route of cell extraction.
The results of Cytopathological examinations consist of reports detailing whether or not irregular cells were found in the cell samples extracted from the patient. If irregular cells were found to be present in the patient cell samples the report will state what type of disease, cancer or infection the patient has and if possible what stage the disease, cancer or infection is currently at. During their examination of cell samples pathologists will centre their focus on the nuclei of the cells and also analyse cell growth, cell metabolism and cell division in order to diagnose presence of illness. The nucleus is the part of the cell containing DNA and RNA and responsible for growth and reproduction.
Cytopathology tests can be used to diagnose autoimmune diseases, cancers, infectious diseases (parasitic, viral, bacterial), amyloidosis, immune system reactions, cell aging, diseases involving certain body cavities, and other illnesses.
Cytopathology AI Research
Towards Artificial Intelligence Applications in Next Generation Cytopathology
https://www.mdpi.com/2227-9059/11/8/2225
Deep Learning for Computational Cytology: A Survey
https://arxiv.org/pdf/2202.05126.pdf
Artificial intelligence and computational pathology
https://www.nature.com/articles/s41374-020-00514-0#Bib1
Deep learning based digital cell profiles for risk stratification of urine cytology images
https://onlinelibrary.wiley.com/doi/full/10.1002/cyto.a.24313
Recent Application of Artificial Intelligence in Non-Gynecological Cancer Cytopathology: A Systematic Review
https://www.mdpi.com/2072-6694/14/14/3529
Advances in Imaging Modalities, Artificial Intelligence, and Single Cell Biomarker Analysis, and Their Applications in Cytopathology
https://www.frontiersin.org/articles/10.3389/fmed.2021.689954/full
A machine learning model for screening of body fluid cytology smears
https://www.biorxiv.org/content/10.1101/2021.07.20.453010v2.full
Automating the Paris System for urine cytopathology—A hybrid deep-learning and morphometric approach
https://acsjournals.onlinelibrary.wiley.com/doi/full/10.1002/cncy.22099
Current status of machine learning in thyroid cytopathology
https://pubmed.ncbi.nlm.nih.gov/37077698/
Fast Pre-Diagnosis of Neoplastic Changes in Cytology Images Using Machine Learning
https://www.mdpi.com/2076-3417/11/16/7181
A Deep Learning Model for Cervical Cancer Screening on Liquid-Based Cytology Specimens in Whole Slide Images
https://www.mdpi.com/2072-6694/14/5/1159
Applications of machine and deep learning to thyroid cytology and histopathology: a review
https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.958310/full
Machine learning-based gene alteration prediction model for primary lung cancer using cytologic images
https://acsjournals.onlinelibrary.wiley.com/doi/full/10.1002/cncy.22609
Quantitative analysis of abnormalities in gynecologic cytopathology with deep learning
https://www.nature.com/articles/s41374-021-00537-1
A Survey for Cervical Cytopathology Image Analysis Using Deep Learning
https://ieeexplore.ieee.org/abstract/document/9046839
A deep learning-based segmentation system for rapid onsite cytologic pathology evaluation of pancreatic masses: A retrospective, multicenter, diagnostic study
https://www.thelancet.com/journals/ebiom/article/PIIS2352-3964(22)00206-7/fulltext
Current status of machine learning in thyroid cytopathology
https://www.sciencedirect.com/science/article/pii/S2153353923001232
Use of Machine Learning–Based Software for the Screening of Thyroid Cytopathology Whole Slide Images
“SMART” cytology: The next generation cytology for precision diagnosis
https://www.sciencedirect.com/science/article/abs/pii/S0740257023000011
Application and performance of artificial intelligence technology in cytopathology
https://www.sciencedirect.com/science/article/abs/pii/S0065128122000496
Molecular Pathology
Molecular Pathology is the diagnosis and study of illness through the microscopic examination of tissues, bodily fluids and organs at the molecular level. Molecular level microscopy enables examination of molecules below the cellular level within samples from the patients anatomy. The term molecule refers to one particle of a substance where that particle contains all the minimum required atoms of elements that constitute the properties of that substance. For example one molecule of Oxygen consists of 2 Oxygen atoms, or one molecule of Water (H2O) consists of 2 Hydrogen atoms and 1 Oxygen atom. Molecular Pathology is concerned with developing and deploying a collection of molecular techniques that analyse diagnostic, prognostic and predictive (treatment) biomarkers in cell genome and proteome in order to diagnose illness, customise treatment for illness and monitor illness/treatment progression. Different Molecular Pathology techniques include; DNA sequencing, In Situ Hybridization, Polymerase Chain Reaction, DNA Microarray, Molecular Pathogen analysis, In Situ RNA sequencing, Immunofluorescence Assays.
Further specific Molecular Pathology diagnostic methods include Specific High Sensitivity Enzymatic Reporter Unlocking (SHERLOCK) which utilises CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) Cas enzymology to detect DNA or RNA sequences that diagnose viruses in patients. Molecular diagnostic methods also include Metagenomic next-generation sequencing (mNGS) which is utilised to diagnose infectious disease in patients by examining the patients nucleic acids (DNA, RNA).
DNA Sequencing refers to the methods of determining the order sequence of nucleotides (thymine, guanine, adenine, cytosine) in one molecule of patient DNA. The full sequence of DNA is known as the human genome which consists of 3 billion base pairs of the above nucleotides. The nucleotide base pairs are copied and transferred with every cell reproduction in the human body. The knowledge of the nucleotide sequence informs pathologists as to whether sections of DNA are regulatory or gene containing, whether changes to said genes could produce disease, and also whether or not the patient is suffering from disease. Genome Sequencing can be used to study and map the genetic characteristics of diseases, develop and improve treatment for these diseases and also monitor treatment/disease progression. For example Genome Sequencing has been used to classify and analyse viruses present in DNA or RNA. The identification of viral genomes has led to over 2 million virus genome sequences being accessible in GenBank. Genetic Sequencing is utilised to study patient genomes in order to diagnose presence of said viruses. The DNA Sequencing methods ardently employed by pathologists include Maxam-Gilbert Sequencing, Frederick Sanger Sequencing, Sequencing By Synthesis, Shotgun Sequencing, Illumina (Solexa) Sequencing, Polony Sequencing, Single Molecule Real Time (SMRT) Sequencing, Nanopore DNA Sequencing.
DNA Sequencing AI Research
A deep learning model for predicting next-generation sequencing depth from DNA sequence
https://www.nature.com/articles/s41467-021-24497-8#Bib1
Review on the Application of Machine Learning Algorithms in the Sequence Data Mining of DNA
https://www.frontiersin.org/articles/10.3389/fbioe.2020.01032/full
A review of deep learning applications in human genomics using next-generation sequencing data
https://humgenomics.biomedcentral.com/articles/10.1186/s40246-022-00396-x
Analysis of DNA Sequence Classification Using CNN and Hybrid Models
https://www.hindawi.com/journals/cmmm/2021/1835056/
Machine Learning Aided Interpretable Approach for Single Nucleotide-Based DNA Sequencing using a Model Nanopore
https://pubs.acs.org/doi/10.1021/acs.jpclett.2c02824
Deep DNA machine learning model to calssify the tumor genome of patients with tumor sequencing
Machine learning for classifying tuberculosis drug-resistance from DNA sequencing data
https://academic.oup.com/bioinformatics/article/34/10/1666/4731737
Beyond sequencing: machine learning algorithms extract biology hidden in Nanopore signal data
https://www.cell.com/trends/genetics/fulltext/S0168-9525(21)00257-2
BioSeq-Analysis: a platform for DNA, RNA and protein sequence analysis based on machine learning approaches
https://academic.oup.com/bib/article/20/4/1280/4763667
Optimizing classification efficiency with machine learning techniques for pattern matching
https://journalofbigdata.springeropen.com/articles/10.1186/s40537-023-00804-6
In Situ Hybridisation is a Molecular Pathology method that involves determining identification and exact location of precise nucleic acid sequences of DNA or RNA within tissue sample cells by introducing probes of external modified nucleic acid strands to the tissue sample. The tissue sample is placed on a slide then the dyed nucleic acid probe is introduced to the tissue and the probe will locate and attach itself to a matching nucleic acid sequence within the tissue. After this binding, the probe and matching nucleic acid sequence can then be observed under microscopic examination revealing the exact location of the targeted nucleic acid sequence within cells. The microscopic examination is used to observe and study the targeted nucleic acid sequence in the context of cellular structure and the context of proteins products from the target gene.
In Situ Hybridisation is utilised to study; genome mapping, morphology and population structure of microorganisms, abnormal gene expression, cytogenetics, viral infection, pathogen profiling, and prenatal diagnosis. Different methods of carrying out In Situ Hybridisation include In Situ Hybridisation (ISH) combined with Polymerase Chain reaction, ISH combined with Catalysed Reporter Deposition (CARD), Fluorescence In Situ Hybridisation (FISH), Multicolour FISH.
In Situ Hybridisation AI Research
Evaluation of deep convolutional neural networks for in situ hybridization gene expression image representation
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0262717
SpotLearn: Convolutional Neural Network for Detection of Fluorescence In Situ Hybridization (FISH) Signals in High-Throughput Imaging Approaches
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6350914/
Deep learning to automatically evaluate HER2 gene amplification status from fluorescence in situ hybridization images
https://www.nature.com/articles/s41598-023-36811-z
DeepSpot: A deep neural network for RNA spot enhancement in single-molecule fluorescence in-situ hybridization microscopy images
DeepBrain: Functional Representation of Neural In-Situ Hybridization Images for Gene Ontology Classification Using Deep Convolutional Autoencoders
https://arxiv.org/abs/1711.09663
An efficient fluorescence in situ hybridization (FISH)-based circulating genetically abnormal cells (CACs) identification method based on Multi-scale MobileNet-YOLO-V4
https://qims.amegroups.org/article/view/90981/html
Phenotyping the virulence of SARS-CoV-2 variants in hamsters by digital pathology and machine learning
https://journals.plos.org/plospathogens/article?id=10.1371/journal.ppat.1011589
Automatic quantification of HER2 gene amplification in invasive breast cancer from chromogenic in situ hybridization whole slide images
A Soft Label Deep Learning to Assist Breast Cancer Target Therapy and Thyroid Cancer Diagnosis
https://www.mdpi.com/2072-6694/14/21/5312
Polymerase Chain Reaction (PCR) is a molecular biology technique that produces millions or even billions of copies of a specific targeted section of a patients DNA that contains a particular sequence of nucleotides (the structural units of DNA) that will then be examined after PCR. Polymerase Chain Reaction is essentially a very extensive exponential amplification of a minimal patient DNA sample (the smallest DNA sample size that can be amplified utilising PCR is one molecule of DNA). These specific DNA sample section copies are needed in aforementioned vast numerical quantities in order to complete various biomedical research procedures.
Polymerase Chain Reaction requires 5 main components in order to reproduce copies of the DNA sample section; including (1) Two sequences of nucleotides from the start and the end of the double stranded DNA sample section being targeted (one sequence from each strand). (2) These two sequences are then used to create a Forward Primer and a Reverse Primer which are synthetic DNA sections (known as oligonucleotides) that correspond to the two sequences of nucleotides chosen at either end of the DNA section sample. The Forward and Reverse Primers will be utilised to bind (anneal) to their corresponding nucleotide sections at either end of the target DNA sample section (thus defining and creating the DNA section to be copied by PCR) and initiate the Polymerase Chain Reaction. (3) DNA Nucleotide bases Adenine, Cytosine, Guanine, and Thymine (Deoxynucleoside Triphosphates or dNTPs) are needed to reproduce the new copy DNA strand as DNA Polymerase uses the dNTPs as foundations for new strand synthesis. (4) DNA Polymerase is an enzyme responsible for synthesising new DNA strands from the instructions provided by the primer prepared target DNA sample section sequence (or instructions from preexisting DNA strands outside of PCR) by sequentially adding available individual DNA Nucleotide bases (dNTPs).
In PCR, Taq (Thermis Aquaticus) DNA Polymerase is the most commonly utilised polymerase due to its heat resistance capability during the denaturation stage, however PCR also alternatively utilises Pfu (Pyrococcus Furiosus) DNA Polymerase due to its replication accuracy. (5) A Buffer chemical solution (liquid solution) contained within a test tube that provides the necessary chemical conditions for the Polymerase Chain Reaction to occur. The test tube and buffer solution containing the PCR components are placed into a thermal cycler which then conducts the stages of the PCR.
The process of carrying out Polymerase Chain Reaction consists of 3 stages; (1) Denaturation, (2) Annealing, (3) Extending. (1) Denaturation involves heating the test tube and PCR components to 95 degrees Celsius which melts the double stranded DNA sample producing division into two separate strands after breaking the hydrogen bonds connecting the bases of the strands. (2) Annealing then involves reducing the temperature of the test tube and solution to 50-65 degrees Celsius in order to allow the Forward and Reverse primers (primers contain 20-30 nucleotide bases) to bind themselves to their corresponding nucleotide sequences that denote the start and end of the target DNA section on the separated DNA strands. The primers added to the 5' end of the separated strands will instruct the DNA polymerase to complete the strands in an opposite direction stopping at the other 3' end. (3) The Extending stage involves configuring the thermal cycler to increase the temperature of the test tube to 72 degrees Celsius. After the primers have bound themselves to the corresponding start/end point nucleotide sequences on the separated strands (thus creating two new separate but incomplete double stranded DNA samples) then the Polymerase enzyme attaches itself to the primers and will initiate the addition of the free DNA Nucleotide bases therefore completing 2 new separate double stranded DNA sample copies.
The Polymerase enzyme will only initiate nucleotide base addition after the primers have bound themselves to the target section as the Polymerase enzymes only add nucleotide bases to double stranded DNA (whether or not the double strands are incomplete is irrelevant).
In the context of medical diagnostics, PCR can be utilised for the following; the preparation of patient DNA for sequencing, prenatal diagnostics, genetically identifying infectious disease pathogens in patients, tissue typing for organ transplantation, the diagnosis of cancer through examining genetic mutations, the individualisation or customisation of cancer treatment for patients.
Polymerase Chain Reaction AI Research
Optimizing polymerase chain reaction (PCR) using machine learning
https://www.biorxiv.org/content/10.1101/2021.08.12.455589v1.full
Anomaly Identification during Polymerase Chain Reaction for Detecting SARS-CoV-2 Using Artificial Intelligence Trained from Simulated Data
https://www.mdpi.com/1420-3049/26/1/20
Deep learning enables accurate analysis of images generated from droplet-based digital polymerase chain reaction (dPCR)
https://www.sciencedirect.com/science/article/abs/pii/S0925400522018846
Quantitative Methylation-Specific Polymerase Chain Reaction Gene Patterns in Urine Sediment Distinguish Prostate Cancer Patients From Control Subjects
Development of machine learning models to predict RT-PCR results for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in patients with influenza-like symptoms using only basic clinical data
https://sjtrem.biomedcentral.com/articles/10.1186/s13049-020-00808-8
Molecular Pathogen Analysis methods include molecular testing methods that examine at the molecular level the antigens or proteins that cause patient illness. Molecular Pathogen analysis include methods such as the Immunochemical assay Enzyme-linked immunosorbent assay (ELISA) (the details of ELISA can be found in the Hematology section above) or the Lateral Flow test (LFT).
The Lateral Flow test performed as an immunochromatographic assay has become widely utilised from laboratories to consumers as a rapid test for detecting the presence of a targeted antigen in a liquid sample such as blood, saliva, serum, or urine (for example pregnancy tests are Lateral Flow tests). Lateral Flow tests consists of placing a liquid sample onto plastic devices within which are sample pads, conjugate pads and nitrocellulose membranes which are absorbent pads made from porous paper which can transport the liquid sample through freeze dried chemical reagents and antibodies. Liquid samples are first placed on the sample pad which filters unneeded liquid sample components. The sample pad then transports the liquid sample to the conjugate pad which contains chemical reagents and coloured antibodies which attach themselves to target antigens therefore labelling any target antigen presence. If the coloured antibody attaches to the target antigen in the liquid sample, they will both be transported past a test line in the nitrocellulose membrane which then utilises further binding molecules to visually indicate the presence of the target antigen and coloured antibody within the liquid sample. Lateral Flow testing can be utilised to examine pregnancy, infectious disease pathogens (viruses, bacteria, fungi, parasites), environmental pollutants, allergens, food contaminants.
ELISA Testing AI Research
Deep learning-assisted ultra-accurate smartphone testing of paper-based colorimetric ELISA assays
https://www.sciencedirect.com/science/article/abs/pii/S0003267023000892
Pixel-Based Machine Learning and Image Reconstitution for Dot-ELISA Pathogen Diagnosis in Biological Samples
https://www.frontiersin.org/articles/10.3389/fmicb.2021.562199/full
Serum markers improve current prediction of metastasis development in early-stage melanoma patients: a machine learning-based study
https://febs.onlinelibrary.wiley.com/doi/full/10.1002/1878-0261.12732
Machine learning-based cytokine microarray digital immunoassay analysis
https://www.sciencedirect.com/science/article/pii/S0956566321001251
Pathological test type and chemical detection using deep neural networks: a case study using ELISA and LFA assays
https://www.emerald.com/insight/content/doi/10.1108/JEIM-01-2020-0038/full/html
Lateral Flow Test AI Research
Machine learning to support visual auditing of home-based lateral flow immunoassay self-test results for SARS-CoV-2 antibodies
https://www.nature.com/articles/s43856-022-00146-z
Machine learning for determining lateral flow device results for testing of SARS-CoV-2 infection in asymptomatic populations
https://www.cell.com/cell-reports-medicine/pdf/S2666-3791(22)00339-1.pdf
SERS-based lateral flow assay combined with machine learning for highly sensitive quantitative analysis of Escherichia coli O157:H7
https://link.springer.com/article/10.1007/s00216-020-02921-0
Deep learning on lateral flow immunoassay for the analysis of detection data
https://www.frontiersin.org/articles/10.3389/fncom.2023.1091180/full
Lateral Flow Assay: A Summary of Recent Progress for Improving Assay Performance
https://www.mdpi.com/2079-6374/13/9/837