Search

An increasing number of machine learning applications are being developed and applied to digital pathology, including hematopathology. The goal of these modern computerized tools is often to support diagnostic workflows by extracting and summarizing information from multiple data sources, including digital images of human tissue. Hematopathology is inherently multimodal and can serve as an ideal case study for machine learning applications. However, hematopathology also poses unique challenges compared to other pathology subspecialities when applying machine learning approaches. By modeling the pathologist workflow and thinking process, machine learning algorithms may be designed to address practical and tangible problems in hematopathology. In this article, we discuss the current trends in machine learning in hematopathology. We review currently available machine learning enabled medical devices supporting hematopathology workflows. We then explore current machine learning research trends of the field with a focus on bone marrow cytology and histopathology, and how adoption of new machine learning tools may be enabled through the transition to digital pathology., (© 2023 The Authors. International Journal of Laboratory Hematology published by John Wiley & Sons Ltd.)

Introduction: Management of hemophilia A has changed significantly in the past few years with the expansion of new and/or modified products as treatment options. Unfortunately, many of the standard factor VIII assays do not always accurately measure all available treatment products; therefore, the laboratory must investigate various assay algorithms to ensure the reporting of the correct results. Methods: Requirements for factor testing, diagnosis and severity levels, product testing, factor VIII inhibitor detection and titers, are evaluated, and potential algorithms are created for optimal assessment of patients with hemophilia A. Results: The potential for inaccurate result reporting for patients with hemophilia A or those being treated with the myriad of products has left many laboratories uncertain as to which assay algorithm to implement to ensure reporting the correct results for all products used in their hemophilia program. Algorithms for using either One‐stage Clotting assays or Chromogenic assays or a combination of both types of assays are presented for each laboratory to implement based on their clinical situation. Conclusions: Several algorithms are considered based on the needs of the clinical providers and their patients. Each laboratory must select a testing algorithm that is cost‐effective and within available resources, yet that encompasses the needs of their providers and patients. Laboratory personnel must consider all assay uses (factor VIII levels, different products, interfering products, and inhibitor titers) in determining the best algorithm for their laboratory. This paper is a starting guide for developing the best factor VIII testing assays and protocols for your laboratory. [ABSTRACT FROM AUTHOR]

Introduction: A peripheral blood smear is a basic test for hematological disease diagnosis. This test is performed manually in many places worldwide, which requires both time and qualified staff. Large laboratories are equipped with digital morphology analyzers, some of which are based on deep learning methods. However, it is difficult to explain to scientists how they work. In this paper, we proposed to add an explanatory factor to enhance the interpretability of deep learning models in leukocyte classification. Methods: 10 297 single images of leukocytes obtained from peripheral blood smears were included in this study. Pre‐trained and fully trained VGG16 and VGG19 models were used to classify the leukocytes, and Shapley Additive Explanations (SHAP) DeepExplainer was applied to visualize the area of cells that were significant for classification. The output images from the DeepExplainer were compared with cellular elements that are essential to laboratory practice. Results: The accuracy of our fully trained models was 99.81% for VGG16 and 99.79% for VGG19. It achieved slightly better results than the partially trained model, which scored 98.67% for VGG16 and 98.33% for VGG19. Their SHAP explanations indicated the significance of cellular structures in microscopic examination. Explanations in the pre‐trained models have proved the cell and nucleus contours to be relevant to classification, while explanations in the fully trained models pointed to the cytoplasm area. Conclusion: Despite different SHAP DeepExplainer explanations for fully and partially trained models, this method appears to be helpful for the verification of leukocyte classification in automated peripheral blood smear examination. [ABSTRACT FROM AUTHOR]

Introduction: Bacteraemia in pregnancy and the post‐partum period can lead to maternal and newborn morbidly. The purpose of this study was to use machine learning tools to identify if bacteraemia in pregnant or post‐partum women could be predicted by full blood count (FBC) parameters other than the white cell count. Methods: The study was performed on 129 women with a positive blood culture (BC) for a clinically significant organism, who had a FBC taken at the same time. They were matched with controls who had a negative BC taken at the same time as a FBC. The data were split in to a training (70%) and test (30%) data set. Machine learning techniques such as recursive partitioning and classification and regression trees were used. Results: A neutrophil/lymphocyte ratio (NLR) of >20 was found to be the most clinically relevant and interpretable construct of the FBC result to predict bacteraemia. The diagnostic accuracy of NLR >20 to predict bacteraemia was then examined. Thirty‐six of the 129 bacteraemia patients had a NLR >20, while only 223 of the 3830 controls had a NLR >20. This gave a sensitivity of 27.9% (95% CI 20.3‐36.4), specificity of 94.1% (93.3‐94.8), positive predictive value of 13.9% (10.6‐17.9) and a negative predictive value (NPV) of 97.4% (97.2‐97.7) when the prevalence of bacteraemia was 3%. Conclusion: The NLR should be considered for use in routine clinical practice when assessing the FBC result in patients with suspected bacteraemia during pregnancy or in the post‐partum period. [ABSTRACT FROM AUTHOR]

Introduction: Coronavirus disease 2019 (COVID‐19) is characterized by a high contagiousness requiring isolation measures. At this time, diagnosis is based on the positivity of specific RT‐PCR and/or chest computed tomography scan, which are time‐consuming and may delay diagnosis. Complete blood count (CBC) can potentially contribute to the diagnosis of COVID‐19. We studied whether the analysis of cellular population data (CPD), provided as part of CBC‐Diff analysis by the DxH 800 analyzers (Beckman Coulter), can help to identify SARS‐CoV‐2 infection. Methods: Cellular population data of the different leukocyte subpopulations were analyzed in 137 controls, 322 patients with proven COVID‐19 (COVID+), and 285 patients for whom investigations were negative for SARS‐CoV‐2 infection (COVID−). When CPD of COVID+ were different from controls and COVID− patients, we used receiver operating characteristic analysis to test the discriminating capacity of the individual parameters. Using a random forest classifier, we developed the algorithm based on the combination of 4 monocyte CPD to discriminate COVID+ from COVID− patients. This algorithm was tested prospectively in a series of 222 patients referred to the emergency unit. Results: Among the 222 patients, 86 were diagnosed as COVID‐19 and 60.5% were correctly identified using the discriminating protocol. Among the 136 COVID− patients, 10.3% were misclassified (specificity 89.7%, sensitivity 60.5%). False negatives were observed mainly in patients with a low inflammatory state whereas false positives were mainly seen in patients with sepsis. Conclusion: Consideration of CPD could constitute a first step and potentially aid in the early diagnosis of COVID‐19. [ABSTRACT FROM AUTHOR]

Introduction: Monocytosis is a frequent trigger for blood smear review in a routine hematology laboratory whereas chronic myelomonocytic leukemia (CMML) is infrequent and arises mostly in elderly patients. In order to define the best workflow for monocytosis, we studied three diagnostic approaches: the classical morphology approach (blood smear review), the flow cytometry assay (quantification of monocyte subsets as described by Selimoglu‐Buet et al in 2015), and the "mono‐dysplasia‐score" also referred to as "Monoscore (as described by our team in 2018 using the structural parameters of the Sysmex XN™ analyzers). Methods: Studying a multicentric cohort of 196 nonclonal monocytoses and CMML patients aged over 50 years, we compared the diagnostic performance of the three approaches alone and in combination to propose a diagnostic decision tree. Results: In patients presenting with additional criteria for slide review to monocytosis (37% of our cohort), we propose to sequentially combine morphology, Monoscore, and flow cytometry. On the contrary, for patients with isolated monocytosis (63%), slide review is not mandatory and we suggest performing flow cytometry depending on the Monoscore value. Using the proposed algorithm, 98% of CMML patients would have been correctly identified, slide review rate drastically reduced, and flow cytometry would have been carried out in 44% of patients. Conclusion: We have shown that implementation of Monoscore is a useful input filter to significantly reduce slide reviews without losing sensitivity and that flow cytometry is a performant technique in the second step of the diagnostic workup of CMML. [ABSTRACT FROM AUTHOR]

Machine learning (ML) offers opportunities to advance pathological diagnosis, especially with increasing trends in digitalizing microscopic images. Diagnosing leukemia is time‐consuming and challenging in many areas globally and there is a growing trend in utilizing ML techniques for its diagnosis. In this review, we aimed to describe the literature of ML utilization in the diagnosis of the four common types of leukemia: acute lymphocytic leukemia (ALL), chronic lymphocytic leukemia (CLL), acute myeloid leukemia (AML), and chronic myelogenous leukemia (CML). Using a strict selection criterion, utilizing MeSH terminology and Boolean logic, an electronic search of MEDLINE and IEEE Xplore Digital Library was performed. The electronic search was complemented by handsearching of references of related studies and the top results of Google Scholar. The full texts of 58 articles were reviewed, out of which, 22 studies were included. The number of studies discussing ALL, AML, CLL, and CML was 12, 8, 3, and 1, respectively. No studies were prospectively applying algorithms in real‐world scenarios. Majority of studies had small and homogenous samples and used supervised learning for classification tasks. 91% of the studies were performed after 2010, and 74% of the included studies applied ML algorithms to microscopic diagnosis of leukemia. The included studies illustrated the need to develop the field of ML research, including the transformation from solely designing algorithms to practically applying them clinically. [ABSTRACT FROM AUTHOR]

Introduction: Morphological assessment of the blood smear has been performed by conventional manual microscopy for many decades. Recently, rapid progress in digital imaging and information technology has led to the development of automated methods of digital morphological analysis of blood smears. Methods: A panel of experts in laboratory hematology reviewed the literature on the use of digital imaging and other strategies for the morphological analysis of blood smears. The strengths and weaknesses of digital imaging were determined, and recommendations on improvement were proposed. Results: By preclassifying cells using artificial intelligence algorithms, digital image analysis automates the blood smear review process and enables faster slide reviews. Digital image analyzers also allow remote networked laboratories to transfer images rapidly to a central laboratory for review, and facilitate a variety of essential work functions in laboratory hematology such as consultations, digital image archival, libraries, quality assurance, competency assessment, education, and training. Different instruments from several manufacturers are available, but there is a lack of standardization of staining methods, optical magnifications, color and display characteristics, hardware, software, and file formats. Conclusion: In order to realize the full potential of Digital Morphology Hematology Analyzers, pre‐analytic, analytic, and postanalytic parameters should be standardized. Manufacturers of new instruments should focus on improving the accuracy of cell preclassifications, and the automated recognition and classification of pathological cell types. Cutoffs for grading morphological abnormalities should depend on clinical significance. With all current devices, a skilled morphologist remains essential for cell reclassification and diagnostic interpretation of the blood smear. [ABSTRACT FROM AUTHOR]

Vitamin K antagonists (VKAs) cannot be administered without regular monitoring in order to assure their efficacy and safety. Indeed, if well managed, the VKAs appear to be no less efficacious or safe than the newer direct oral anticoagulants (DOACs). Although it is claimed that no regular monitoring of the DOACs is needed, their levels are increasingly being measured under a variety of circumstances, for example, prior to surgery, in suspected overdose, to confirm effective reversal, in patients with malabsorption and to assess patient compliance. Although no therapeutic range has been identified for the DOACs, it has been demonstrated for dabigatran and edoxaban that their antithrombotic effect increases gradually with increasing concentrations and that the risk of major bleeding also gradually increases. Furthermore, it has been determined that almost all dabigatran‐related thrombotic events occur in patients with the lowest quartile concentration of the drug. This suggests that to assure an ideal effect of DOACs in all patients taking them, some form of regular monitoring and dose tailoring should be performed. For the vitamin K antagonists, the best outcome is obtained using formal algorithms and centralized management. Furthermore, data suggest that replacing the standard prothrombin time as a monitoring test may increase the stability of VKA anticoagulation with consequent reduction in thromboembolism without an increase in bleeding. Thus, it is likely that the outcome of all current oral anticoagulants can be improved in the coming years by improving monitoring and tailoring their effect. [ABSTRACT FROM AUTHOR]

Introduction The use of adapted cut-off values in the elderly, combined with clinical probability ( PTP), increases the proportion of patients in whom venous thromboembolism ( VTE) can be safely excluded, compared with the conventional cut-off value of 500 μg/L fibrinogen equivalent units ( FEU). We evaluated the clinical performance of three different approaches to establish cut-off values for a D-dimer assay whose results are expressed in D-dimer units (D- DU). Methods Hemos IL D-dimer HS assay (Instrumentation Laboratory) was performed in 279 consecutive outpatients with suspected deep venous thrombosis ( DVT) and nonhigh PTP. Results Considering patients >60 years, the number of negative D-dimer results increased using the modified (376 ng/mL if ≥60 years) and the age-adjusted cut-off (age years × 5 ng/mL if >50 years) compared to the conventional one (230 ng/mL for all patients; 54.6%, 58.2%, and 25.0%, respectively), with no false-negative results. The higher increase was observed in patients >80 years (43.9%, 56.1%, and 8.8%, respectively). Conclusion For the Hemos IL D-dimer HS, the use of specific cut-off values in older subjects with suspected DVT and nonhigh PTP increases the number of patients in whom DVT can be safely excluded. [ABSTRACT FROM AUTHOR]

Introduction: Thalassemia and iron deficiency may both result in hypochromic microcytic anemia. Hematological algorithms that differentiate the two are mainly established in adult selected diagnostic groups. We aimed at creating an algorithm applicable in the presence of children, hemoglobin variants, and iron deficiency. Methods: Our study material constituted blood samples referred during 1 year for routine diagnostics of hemoglobinopathy. We included 443 samples, of which 37% were from children 3 months or older. We found β-thalassemia trait (n = 100), α -thalassemia (n = 75), combined α -/β-thalassemia (n = 14), hemoglobin variants (n = 42), and no-hemoglobinopathy (n = 207), of whom 107 had a ferritin at or below 20 µg/L. We included reticulocyte hemoglobin equivalent, ferritin, and erythrocyte count in our algorithm. Results: Our algorithm differentiated β-thalassemia trait from no-hemoglobinopathy with a sensitivity of 99% at 83% specificity. It performed better than other published algorithms when applied to all patient samples, while equally or moderately better in the 63% adult samples. Our algorithm also detected the clinically significant ot-thalassemias, and most of the combined α /β thalassemias and thalassemic hemoglobin variants. Conclusion: Our algorithm efficiently differentiated thalassemia and thalassemic hemoglobin variants from iron deficiency in children and adults. [ABSTRACT FROM AUTHOR]