Multiparameter interpretative reporting in diagnostic laboratory hematology

Artificial Intelligence-Assisted Diagnostic Cytology and Genomic Testing for Hematologic Disorders

Artificial intelligence (AI) is a rapidly evolving field of computer science that involves the development of computational programs that can mimic human intelligence. In particular, machine learning and deep learning models have enabled the identification and grouping of patterns within data, leading to the development of AI systems that have been applied in various areas of hematology, including digital pathology, alpha thalassemia patient screening, cytogenetics, immunophenotyping, and sequencing. These AI-assisted methods have shown promise in improving diagnostic accuracy and efficiency, identifying novel biomarkers, and predicting treatment outcomes. However, limitations such as limited databases, lack of validation and standardization, systematic errors, and bias prevent AI from completely replacing manual diagnosis in hematology. In addition, the processing of large amounts of patient data and personal information by AI poses potential data privacy issues, necessitating the development of regulations to evaluate AI systems and address ethical concerns in clinical AI systems. Nonetheless, with continued research and development, AI has the potential to revolutionize the field of hematology and improve patient outcomes. To fully realize this potential, however, the challenges facing AI in hematology must be addressed and overcome.

Artificial intelligence: Training the trainer

Including artificial intelligence in haematological education is compulsory but should not be limited to students. Experienced haematologist and decision-makers in the clinical environment have at least similar needs. This is because of the tremendous potential, opportunities and benefits the timely inclusion of artificial intelligence offers in diagnosis, prediction and personalised therapy. Commentary on: Chai et al. Integrating artificial intelligence into haematology training and practice: opportunities, threats and proposed solutions. Br J Hematol 2022. https://doi.org/10.1111/bjh.18343.

Machine Learning Applications in the Diagnosis of Benign and Malignant Hematological Diseases

The use of machine learning (ML) and deep learning (DL) methods in hematology includes diagnostic, prognostic, and therapeutic applications. This increase is due to the improved access to ML and DL tools and the expansion of medical data. The utilization of ML remains limited in clinical practice, with some disciplines further along in their adoption, such as radiology and histopathology. In this review, we discuss the current uses of ML in diagnosis in the field of hematology, including image-recognition, laboratory, and genomics-based diagnosis. Additionally, we provide an introduction to the fields of ML and DL, highlighting current trends, limitations, and possible areas of improvement.

Artificial intelligence in Hematology

Artificial intelligence (AI) is a computer based science which aims to simulate human brain faculties using a computational system. A brief history of this new science goes from the creation of the first artificial neuron in 1943 to the first artificial neural network application to genetic algorithms. The potential for a similar technology in medicine has immediately been identified by scientists and researchers. The possibility to store and process all medical knowledge has made this technology very attractive to assist or even surpass clinicians in reaching a diagnosis. Applications of AI in medicine include devices applied to clinical diagnosis in neurology and cardiopulmonary diseases, as well as the use of expert or knowledge-based systems in routine clinical use for diagnosis, therapeutic management and for prognostic evaluation. Biological applications include genome sequencing or DNA gene expression microarrays, modeling gene networks, analysis and clustering of gene expression data, pattern recognition in DNA and proteins, protein structure prediction. In the field of hematology the first devices based on AI have been applied to the routine laboratory data management. New tools concern the differential diagnosis in specific diseases such as anemias, thalassemias and leukemias, based on neural networks trained with data from peripheral blood analysis. A revolution in cancer diagnosis, including the diagnosis of hematological malignancies, has been the introduction of the first microarray based and bioinformatic approach for molecular diagnosis: a systematic approach based on the monitoring of simultaneous expression of thousands of genes using DNA microarray, independently of previous biological knowledge, analysed using AI devices. Using gene profiling, the traditional diagnostic pathways move from clinical to molecular based diagnostic systems.

Haematological validation of a computer-based bone marrow reporting system

AIMS

To prove the safety and effectiveness of "Professor Belmonte", a knowledge-based system for bone marrow reporting, a formal evaluation of the reports generated by the system was performed.

METHODS

Three haematologists (a consultant, a senior registrar, and a junior registrar), none of whom were involved in the development of the software, compared the unedited reports generated by Professor Belmonte with the original bone marrow reports in 785 unselected cases. Each haematologist independently graded the quality of Belmonte's reports using one of four categories: (a) better than the original report (more informative, containing useful information missing in the original report); (b) equivalent to the original report; (c) satisfactory, but missing information that should have been included; and (d) unsatisfactory.

RESULTS

The consultant graded 64 reports as more informative than the original, 687 as equivalent to the original, 32 as satisfactory, and two as unsatisfactory. The senior registrar considered 29 reports to be better than the original, 739 to be equivalent to the original, 15 to be satisfactory, and two to be unsatisfactory. The junior registrar found that 88 reports were better than the original, 681 were equivalent to the original, 14 were satisfactory, and two were unsatisfactory. Each judge found two different reports to be unsatisfactory according to their criteria. All 785 reports generated by the computer system received at least two scores of satisfactory or better.

CONCLUSIONS

In this representative study, Professor Belmonte generated bone marrow reports that proved to be as accurate as the original reports in a large university hospital. The haematology knowledge contained within the system, the reasoning process, and the function of the software are safe and effective for assisting haematologists in generating high quality bone marrow reports.

Are normative expert systems appropriate for diagnostic pathology?

Conceptual models for diagnostic reasoning proposed in the medical literature are presented to stimulate discussion about the issue of the appropriateness of probabilistic knowledge-based systems for medical diagnosis. Evidence is presented to corroborate the authors' view that diagnosis is a problem-solving task, rather than a decision-making task. In the authors' opinion, probabilistic reasoning is better suited to situations dealing with choices for clinical intervention, rather than to those dealing with determining the correct diagnosis. A critique is given of a diagnostic Bayesian expert system for lymph node pathology. In empirical studies, diagnostic Bayesian systems have been shown to typically list the correct diagnosis as the program's first choice 60% to 70% of the time. One reason for this undistinguished level of diagnostic performance is that Bayesian systems are not designed to represent and use knowledge the same way that an expert does.