Inter-observer variance and the need for standardization in the morphological classification of myelodysplastic syndrome

Deep Learning-Based Blood Abnormalities Detection as a Tool for VEXAS Syndrome Screening

INTRODUCTION

VEXAS is a syndrome described in 2020, caused by mutations of the UBA1 gene, and displaying a large pleomorphic array of clinical and hematological features. Nevertheless, these criteria lack significance to discriminate VEXAS from other inflammatory conditions at the screening step. This work hence first focused on singling out dysplastic features indicative of the syndrome among peripheral blood (PB) polymorphonuclears (PMN). A deep learning algorithm is then proposed for automatic detection of these features.

METHODS

A multicentric dataset, comprising 9514 annotated PMN images was gathered, including UBA1 mutated VEXAS (n = 25), UBA1 wildtype myelodysplastic (n = 14), and UBA1 wildtype cytopenic patients (n = 25). Statistical analysis on a subset of patients was performed to screen for significant abnormalities. Detection of these features on PB was then automated with a convolutional neural network (CNN) for multilabel classification.

RESULTS

Significant differences were observed in the proportions of PMNs with pseudo-Pelger, nuclear spikes, vacuoles, and hypogranularity between patients with VEXAS and both cytopenic and myelodysplastic controls. Automatic detection of these abnormalities yielded AUCs in the range [0.85-0.97] and a F1-score of 0.70 on the test set. A VEXAS screening score was proposed, leveraging the model outputs and predicting the UBA1 mutational status with 0.82 sensitivity and 0.71 specificity on the test patients.

CONCLUSION

This study suggests that computer-assisted analysis of PB smears, focusing on suspected VEXAS cases, can provide valuable insights for determining which patients should undergo molecular testing. The presented deep learning approach can help hematologists direct their suspicions before initiating further analyses.

High-Resolution Raman Imaging of >300 Patient-Derived Cells from Nine Different Leukemia Subtypes: A Global Clustering Approach

Leukemia comprises a diverse group of bone marrow tumors marked by cell proliferation. Current diagnosis involves identifying leukemia subtypes through visual assessment of blood and bone marrow smears, a subjective and time-consuming method. Our study introduces the characterization of different leukemia subtypes using a global clustering approach of Raman hyperspectral maps of cells. We analyzed bone marrow samples from 19 patients, each presenting one of nine distinct leukemia subtypes, by conducting high spatial resolution Raman imaging on 319 cells, generating over 1.3 million spectra in total. An automated preprocessing pipeline followed by a single-step global clustering approach performed over the entire data set identified relevant cellular components (cytoplasm, nucleus, carotenoids, myeloperoxidase (MPO), and hemoglobin (HB)) enabling the unsupervised creation of high-quality pseudostained images at the single-cell level. Furthermore, this approach provided a semiquantitative analysis of cellular component distribution, and multivariate analysis of clustering results revealed the potential of Raman imaging in leukemia research, highlighting both advantages and challenges associated with global clustering.

A comparative evaluation of three consecutive artificial intelligence algorithms released by Techcyte for identification of blasts and white blood cells in abnormal peripheral blood films

INTRODUCTION

Digital pathology artificial intelligence (AI) platforms have the capacity to improve over time through "deep machine learning." We have previously reported on the accuracy of peripheral white blood cell (WBC) differential and blast identification by Techcyte (Techcyte, Inc., Orem, UT, USA), a digital scanner-agnostic web-based system for blood film reporting. The aim of the current study was to compare AI protocols released over time to assess improvement in cell identification.

METHODS

WBC differentials were performed using Techcyte's online AI software on the same 124 digitized abnormal peripheral blood films (including 64 acute and 22 chronic leukaemias) in 2019 (AI1), 2020 (AI2), and 2022 (AI3), with no reassignment by a morphologist at any time point. AI results were correlated to the "gold standard" of manual microscopy, and comparison of Lin's concordance coefficients (LCC) and sensitivity and specificity of blast identification were used to determine the superior AI version.

RESULTS

AI correlations (r) with manual microscopy for individual cell types ranged from 0.50-0.90 (AI1), 0.66-0.86 (AI2) and 0.71-0.91 (AI3). AI3 concordance with manual microscopy was significantly improved compared to AI1 for identification of neutrophils (LCC AI3 = 0.86 vs. AI1 = 0.77, p = 0.03), total granulocytes (LCC AI3 = 0.92 vs. AI1 = 0.82, p = 0.0008), immature granulocytes (LCC AI3 = 0.67 vs. AI1 = 0.38, p = 0.0014), and promyelocytes (LCC AI3 = 0.53 vs. AI1 = 0.16, p = 0.0008). Sensitivity for blast identification (n = 65 slides) improved from 97% (AI1), to 98% (AI2), to 100% (AI3), while blast specificity decreased from 24% (AI1), to 14% (AI2) to 12% (AI3).

CONCLUSION

Techcyte AI has shown significant improvement in cell identification over time and maintains high sensitivity for blast identification in malignant films.

Deep immunophenotypic analysis of the bone marrow progenitor cells in myelodysplastic syndromes

BACKGROUND

Diagnosis of myelodysplastic syndromes (MDS) is often challenging and requires integration of clinical, morphologic, cytogenetics and molecular information. Flow cytometry immunophenotyping (FCIP) can support the diagnosis by demonstration of numerical and immunophenotypic abnormalities of progenitor and maturing myelomonocytic and erythroid populations. We have previously shown that comprehensive immunophenotypic analysis of the progenitor population is valuable in the diagnosis of MDS and myelodysplastic/myeloproliferative neoplasms (MDS/MPN). This study was designed to improve the analysis method and confirm its value in a larger cohort of patients.

METHODS

FCIP of bone marrow samples from 105 patients with cytopenia(s) (with or without leukocytosis) and clinical concern for MDS or MDS/MPN was performed using a single-tube/10-color/13-marker assay. A modified analysis approach was used to obtain 11 progenitor parameters and 2 myelomonocytic parameters.

RESULTS

Significantly higher number of abnormalities were identified in MDS and MDS/MPN cases when compared to cytopenic patients not meeting the diagnostic criteria for MDS (Non-MDS). A FCIP score that combined the 13 parameters showed a sensitivity of 89.8% and specificity of 93.5% for the diagnosis of MDS and MDS/MPN. The sensitivity was 100% for both MDS/MPN and higher-risk MDS, and 81.3% for lower-risk MDS.

CONCLUSION

This study confirms that detailed immunophenotypic analysis of the progenitor population is powerful in the diagnosis of MDS and MDS/MPN. The combination of markers used in the panel allowed for evaluation of two relatively new parameters, namely myeloid progenitor heterogeneity and stem cell aberrancy, which improved the sensitivity of the assay for lower-risk MDS.

Whole slide image representation in bone marrow cytology

One of the goals of AI-based computational pathology is to generate compact representations of whole slide images (WSIs) that capture the essential information needed for diagnosis. While such approaches have been applied to histopathology, few applications have been reported in cytology. Bone marrow aspirate cytology is the basis for key clinical decisions in hematology. However, visual inspection of aspirate specimens is a tedious and complex process subject to variation in interpretation, and hematopathology expertise is scarce. The ability to generate a compact representation of an aspirate specimen may form the basis for clinical decision-support tools in hematology. In this study, we leverage our previously published end-to-end AI-based system for counting and classifying cells from bone marrow aspirate WSIs, which enables the direct use of individual cells as inputs rather than WSI patches. We then construct bags of individual cell features from each WSI, and apply multiple instance learning to extract their vector representations. To evaluate the quality of our representations, we conducted WSI retrieval and classification tasks. Our results show that we achieved a mAP@10 of 0.58 ±0.02 in WSI-level image retrieval, surpassing the random-retrieval baseline of 0.39 ±0.1. Furthermore, we predicted five diagnostic labels for individual aspirate WSIs with a weighted-average F1 score of 0.57 ±0.03 using a k-nearest-neighbors (k-NN) model, outperforming guessing using empirical class prior probabilities (0.26 ±0.02). We present the first example of exploring trainable mechanisms to generate compact, slide-level representations in bone marrow cytology with deep learning. This method has the potential to summarize complex semantic information in WSIs toward improved diagnostics in hematology, and may eventually support AI-assisted computational pathology approaches.

Patterns of Diagnostic Evaluation and Determinants of Treatment in Older Patients With Non-transfusion Dependent Myelodysplastic Syndromes

BACKGROUND

Older patients with myelodysplastic syndromes (MDS), particularly those with no or one cytopenia and no transfusion dependence, typically have an indolent course. Approximately, half of these receive the recommended diagnostic evaluation (DE) for MDS. We explored factors determining DE in these patients and its impact on subsequent treatment and outcomes.

PATIENTS AND METHODS

We used 2011-2014 Medicare data to identify patients ≥66 years of age diagnosed with MDS. We used Classification and Regression Tree (CART) analysis to identify combinations of factors associated with DE and its impact on subsequent treatment. Variables examined included demographics, comorbidities, nursing home status, and investigative procedures performed. We conducted a logistic regression analysis to identify correlates associated with receipt of DE and treatment.

RESULTS

Of 16 851 patients with MDS, 51% underwent DE. patients with MDS with no cytopenia (n = 3908) had the lowest uptake of DE (34.7%). Compared to patients with no cytopenia, those with any cytopenia had nearly 3 times higher odds of receiving DE [adjusted odds ratio (AOR), 2.81: 95% CI, 2.60-3.04] and the odds were higher for men than for women [AOR, 1.39: 95%CI, 1.30-1.48] and for Non-Hispanic Whites [vs. everyone else (AOR, 1.17: 95% CI, 1.06-1.29)]. The CART showed DE as the principal discriminating node, followed by the presence of any cytopenia for receiving MDS treatment. The lowest percentage of treatment was observed in patients without DE, at 14.6%.

CONCLUSION

In this select older patients with MDS, we identified disparities in accurate diagnosis by demographic and clinical factors. Receipt of DE influenced subsequent treatment but not survival.

Cell projection plots: A novel visualization of bone marrow aspirate cytology

Deep models for cell detection have demonstrated utility in bone marrow cytology, showing impressive results in terms of accuracy and computational efficiency. However, these models have yet to be implemented in the clinical diagnostic workflow. Additionally, the metrics used to evaluate cell detection models are not necessarily aligned with clinical goals and targets. In order to address these issues, we introduce novel, automatically generated visual summaries of bone marrow aspirate specimens called cell projection plots (CPPs). Encompassing relevant biological patterns such as neutrophil maturation, CPPs provide a compact summary of bone marrow aspirate cytology. To gauge clinical relevance, CPPs were inspected by 3 hematopathologists, who decided whether corresponding diagnostic synopses matched with generated CPPs. Pathologists were able to match CPPs to the correct synopsis with a matching degree of 85%. Our finding suggests CPPs can represent clinically relevant information from bone marrow aspirate specimens and may be used to efficiently summarize bone marrow cytology to pathologists. CPPs could be a step toward human-centered implementation of artificial intelligence (AI) in hematopathology, and a basis for a diagnostic-support tool for digital pathology workflows.

Digital assessment of peripheral blood and bone marrow aspirate smears

The diagnosis of benign and neoplastic hematologic disorders relies on analysis of peripheral blood and bone marrow aspirate smears. As demonstrated by the widespread laboratory adoption of hematology analyzers for automated assessment of peripheral blood, digital analysis of these samples provides many significant benefits compared to relying solely on manual review. Nonetheless, analogous instruments for digital bone marrow aspirate smear assessment have yet to be clinically implemented. In this review, we first provide a historical overview detailing the implementation of hematology analyzers for digital peripheral blood assessment in the clinical laboratory, including the improvements in accuracy, scope, and throughput of current instruments over prior generations. We also describe recent research in digital peripheral blood assessment, particularly in the development of advanced machine learning models that may soon be incorporated into commercial instruments. Next, we provide an overview of recent research in digital assessment of bone marrow aspirate smears and how these approaches could soon lead to development and clinical adoption of instrumentation for automated bone marrow aspirate smear analysis. Finally, we describe the relative advantages and provide our vision for the future of digital assessment of peripheral blood and bone marrow aspirate smears, including what improvements we can soon expect in the hematology laboratory.

ERCC6L2-related disease: a novel entity of bone marrow failure disorder with high risk of clonal evolution

ERCC excision repair 6 like 2 (ERCC6L2) gene encodes for different helicase-like protein members of the Snf2 family involved in transcription-coupled nucleotide excision repair and in cell proliferation. Germline homozygous mutations in children and adults predispose to a peculiar bone marrow failure phenotype characterized by mild hematological alterations with a high risk of developing acute myeloid leukemia. The outcome for patients with leukemia progression is dismal while patients undergoing hematopoietic stem cell transplantation in the early stage have better outcomes. The ERCC6L2-related hematological disease presents a high penetrance, posing important questions regarding the treatment strategies and possible preemptive approaches. This review describes the biological function of ERCC6L2 and the clinical manifestations of the associated disease, trying to focus on the unsolved clinical questions.

Hematology and Machine Learning

BACKGROUND

Substantial improvements in computational power and machine learning (ML) algorithm development have vastly increased the limits of what autonomous machines are capable of. Since its beginnings in the 19th century, laboratory hematology has absorbed waves of progress yielding improvements in both of accuracy and efficiency. The next wave of change in laboratory hematology will be the result of the ML revolution that has already touched many corners of healthcare and society at large.

CONTENT

This review will describe the manifestations of ML and artificial intelligence (AI) already utilized in the clinical hematology laboratory. This will be followed by a topical summary of the innovative and investigational applications of this technology in each of the major subdomains within laboratory hematology.

SUMMARY

Application of this technology to laboratory hematology will increase standardization and efficiency by reducing laboratory staff involvement in automatable activities. This will unleash time and resources for focus on more meaningful activities such as the complexities of patient care, research and development, and process improvement.

Digital morphology in hematology diagnosis and education: The experience of the European LeukemiaNet WP10

Hematological diagnostics is based on increasingly precise techniques of cellular and molecular analysis. The correct interpretation of the blood and bone marrow smears observed under an optical microscope still represents a cornerstone. Precise quantitative and qualitative cytomorphological criteria have recently been codified by up-to-date guidelines for diagnosing hematopoietic neoplasms. Morphological analysis has found formidable support in digital reproduction techniques, which have simplified the circulation of images for educational or consultation purposes. From 2007 to 2019, the Working Group WP10 of European LeukemiaNet (ELN) used, in annual exercises, digital images to support training in cytomorphology and verify harmonization and comparability in the interpretation of blood and bone marrow smears. We describe the design, development, and results of this program, which had 741 participants in-person or remotely, to which 2055 questions were submitted regarding the interpretation of cytomorphological images. We initially used circulation and presentation of digital microphotographs and then introduced a virtual microscopy (VM). Virtual slides were obtained using a whole slide imaging technique, similar to the one largely used in histopathology, to produce digitized scans of consecutive microscopic fields and reassembles them to obtain a complete virtual smear by stitching. Participants were required to identify cells in labeled fields of view of the virtual slides to obtain a morphological diagnosis. This work has demonstrated substantial improvements in diagnostic accuracy and harmonization with the VM technique. Between-observer concordance increased from 62.5% to 83.0%. The integrity of the digitalized film image, which provides a general context for cell abnormalities, was the main factor for this outcome.

Automated bone marrow cytology using deep learning to generate a histogram of cell types

Background

Bone marrow cytology is required to make a hematological diagnosis, influencing critical clinical decision points in hematology. However, bone marrow cytology is tedious, limited to experienced reference centers and associated with inter-observer variability. This may lead to a delayed or incorrect diagnosis, leaving an unmet need for innovative supporting technologies.

Methods

We develop an end-to-end deep learning-based system for automated bone marrow cytology. Starting with a bone marrow aspirate digital whole slide image, our system rapidly and automatically detects suitable regions for cytology, and subsequently identifies and classifies all bone marrow cells in each region. This collective cytomorphological information is captured in a representation called Histogram of Cell Types (HCT) quantifying bone marrow cell class probability distribution and acting as a cytological patient fingerprint.

Results

Our system achieves high accuracy in region detection (0.97 accuracy and 0.99 ROC AUC), and cell detection and cell classification (0.75 mean average precision, 0.78 average F1-score, Log-average miss rate of 0.31).

Conclusions

HCT has potential to eventually support more efficient and accurate diagnosis in hematology, supporting AI-enabled computational pathology.

A geno-clinical decision model for the diagnosis of myelodysplastic syndromes

We developed a machine learning–based model to assist in the differential diagnosis of myeloid malignancies.

Our work also describes genotype-phenotype correlations in different myeloid malignancies.

The differential diagnosis of myeloid malignancies is challenging and subject to interobserver variability. We used clinical and next-generation sequencing (NGS) data to develop a machine learning model for the diagnosis of myeloid malignancies independent of bone marrow biopsy data based on a 3-institution, international cohort of patients. The model achieves high performance, with model interpretations indicating that it relies on factors similar to those used by clinicians. In addition, we describe associations between NGS findings and clinically important phenotypes and introduce the use of machine learning algorithms to elucidate clinicogenomic relationships.

White blood cell evaluation in haematological malignancies using a web-based digital microscopy platform

INTRODUCTION

Digital microscopy systems are beginning to replace traditional light microscopes for morphologic analysis of blood films, but these are geographically restricted to individual computers and technically limited by manufacturer's constraints. We explored the use of a scanner-agnostic web-based artificial intelligence (AI) system to assess the accuracy of white blood cell (WBC) differentials and blast identification in haematological malignancies.

METHODS

Digitized images of 20 normal and 124 abnormal peripheral blood films were uploaded to the web-based platform (Techcyte©) and WBC differentials performed using the online AI software. Digital images were viewed for accuracy and manual cell reassignment was performed where necessary. Results were correlated to the 'gold standard' of manual microscopy for each WBC class, and sensitivity and specificity of blast identification were calculated.

RESULTS

The AI digital differential was very strongly correlated to microscopy (r > .8) for most normal cell types and did not require any manual reassignment. The AI digital differential was less reliable for abnormal blood films (r = .50-.87), but could be greatly improved by manual assessment of digital images for most cell types (r > .95) with the exception of immature granulocytes (r = .62). For blast identification, initial AI digital differentials showed 96% sensitivity and 25% specificity, which was improved to 99% and 84%, respectively, after manual digital review.

CONCLUSIONS

The Techcyte platform allowed remote viewing and manual analysis of digitized slides that was comparable to microscopy. The AI software produced adequate WBC differentials for normal films and had high sensitivity for blast identification in malignant films.

A new convolutional neural network predictive model for the automatic recognition of hypogranulated neutrophils in myelodysplastic syndromes

BACKGROUND

Dysplastic neutrophils commonly show at least 2/3 reduction of the content of cytoplasmic granules by morphologic examination. Recognition of less granulated dysplastic neutrophils by human eyes is difficult and prone to inter-observer variability. To tackle this problem, we proposed a new deep learning model (DysplasiaNet) able to automatically recognize the presence of hypogranulated dysplastic neutrophils in peripheral blood.

METHODS

Eight models were generated by varying convolutional blocks, number of layer nodes and fully connected layers. Each model was trained for 20 epochs. The five most accurate models were selected for a second stage, being trained again from scratch for 100 epochs. After training, cut-off values were calculated for a granularity score that discerns between normal and dysplastic neutrophils. Furthermore, a threshold value was obtained to quantify the minimum proportion of dysplastic neutrophils in the smear to consider that the patient might have a myelodysplastic syndrome (MDS). The final selected model was the one with the highest accuracy (95.5%).

RESULTS

We performed a final proof of concept with new patients not involved in previous steps. We reported 95.5% sensitivity, 94.3% specificity, 94% precision, and a global accuracy of 94.85%.

CONCLUSIONS

The primary contribution of this work is a predictive model for the automatic recognition in an objective way of hypogranulated neutrophils in peripheral blood smears. We envision the utility of the model implemented as an evaluation tool for MDS diagnosis integrated in the clinical laboratory workflow.

Searching Images for Consensus: Can AI Remove Observer Variability in Pathology?

One of the major obstacles in reaching diagnostic consensus is observer variability. With the recent success of artificial intelligence, particularly the deep networks, the question emerges as to whether the fundamental challenge of diagnostic imaging can now be resolved. This article briefly reviews the problem and how eventually both supervised and unsupervised AI technologies could help to overcome it.

Utility of clinical comprehensive genomic characterization for diagnostic categorization in patients presenting with hypocellular bone marrow failure syndromes

Bone marrow failure (BMF) related to hypoplasia of hematopoietic elements in the bone marrow is a heterogeneous clinical entity with a broad differential diagnosis including both inherited and acquired causes. Accurate diagnostic categorization is critical to optimal patient care and detection of genomic variants in these patients may provide this important diagnostic and prognostic information. We performed real-time, accredited (ISO15189) comprehensive genomic characterization including targeted sequencing and whole exome sequencing in 115 patients with BMF syndrome (median age 24 years, range 3 months - 81 years). In patients with clinical diagnoses of inherited BMF syndromes, acquired BMF syndromes or clinically unclassifiable BMF we detected variants in 52% (12/23), 53% (25/47) and 56% (25/45) respectively. Genomic characterization resulted in a change of diagnosis in 30/115 (26%) including the identification of germline causes for 3/47 and 16/45 cases with pre-test diagnoses of acquired and clinically unclassifiable BMF respectively. The observed clinical impact of accurate diagnostic categorization included choice to perform allogeneic stem cell transplantation, disease-specific targeted treatments, identification of at-risk family members and influence of sibling allogeneic stem cell donor choice. Multiple novel pathogenic variants and copy number changes were identified in our cohort including in TERT, FANCA, RPS7 and SAMD9. Whole exome sequence analysis facilitated the identification of variants in two genes not typically associated with a primary clinical manifestation of BMF but also demonstrated reduced sensitivity for detecting low level acquired variants. In conclusion, genomic characterization can improve diagnostic categorization of patients presenting with hypoplastic BMF syndromes and should be routinely performed in this group of patients.

How I investigate difficult cells at the optical microscope

Blood cell morphological identification on the peripheral blood and bone marrow films remains a cornerstone for the diagnosis of hematological neoplasms to be integrated with immunophenotyping, molecular genetics, and histopathology. Although standardization is still far from being achieved, with high interobserver variability, in recent years, several classification approaches, from the 1976 FAB to the 2016 WHO classification, have provided hematologists with detailed morphological descriptions for a large number of diseases. Counting blasts and detecting dysplastic specimens are two cornerstones of morphological diagnosis. This review deals with identifying difficult cells, with particular reference of those with relevant diagnostic implications.

Detection and Classification of Immature Leukocytes for Diagnosis of Acute Myeloid Leukemia Using Random Forest Algorithm

Acute myeloid leukemia (AML) is a fatal blood cancer that progresses rapidly and hinders the function of blood cells and the immune system. The current AML diagnostic method, a manual examination of the peripheral blood smear, is time consuming, labor intensive, and suffers from considerable inter-observer variation. Herein, a machine learning model to detect and classify immature leukocytes for efficient diagnosis of AML is presented. Images of leukocytes in AML patients and healthy controls were obtained from a publicly available dataset in The Cancer Imaging Archive. Image format conversion, multi-Otsu thresholding, and morphological operations were used for segmentation of the nucleus and cytoplasm. From each image, 16 features were extracted, two of which are new nucleus color features proposed in this study. A random forest algorithm was trained for the detection and classification of immature leukocytes. The model achieved 92.99% accuracy for detection and 93.45% accuracy for classification of immature leukocytes into four types. Precision values for each class were above 65%, which is an improvement on the current state of art. Based on Gini importance, the nucleus to cytoplasm area ratio was a discriminative feature for both detection and classification, while the two proposed features were shown to be significant for classification. The proposed model can be used as a support tool for the diagnosis of AML, and the features calculated to be most important serve as a baseline for future research.

Assessment of dysplasia in bone marrow smear with convolutional neural network

In this study, we developed the world's first artificial intelligence (AI) system that assesses the dysplasia of blood cells on bone marrow smears and presents the result of AI prediction for one of the most representative dysplasia—decreased granules (DG). We photographed field images from the bone marrow smears from patients with myelodysplastic syndrome (MDS) or non-MDS diseases and cropped each cell using an originally developed cell detector. Two morphologists labelled each cell. The degree of dysplasia was evaluated on a four-point scale: 0–3 (e.g., neutrophil with severely decreased granules were labelled DG3). We then constructed the classifier from the dataset of labelled images. The detector and classifier were based on a deep neural network pre-trained with natural images. We obtained 1797 labelled images, and the morphologists determined 134 DGs (DG1: 46, DG2: 77, DG3: 11). Subsequently, we performed a five-fold cross-validation to evaluate the performance of the classifier. For DG1–3 labelled by morphologists, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were 91.0%, 97.7%, 76.3%, 99.3%, and 97.2%, respectively. When DG1 was excluded in the process, the sensitivity, specificity, PPV, NPV, and accuracy were 85.2%, 98.9%, 80.6%, and 99.2% and 98.2%, respectively.

Pan-cancer diagnostic consensus through searching archival histopathology images using artificial intelligence

The emergence of digital pathology has opened new horizons for histopathology. Artificial intelligence (AI) algorithms are able to operate on digitized slides to assist pathologists with different tasks. Whereas AI-involving classification and segmentation methods have obvious benefits for image analysis, image search represents a fundamental shift in computational pathology. Matching the pathology of new patients with already diagnosed and curated cases offers pathologists a new approach to improve diagnostic accuracy through visual inspection of similar cases and computational majority vote for consensus building. In this study, we report the results from searching the largest public repository (The Cancer Genome Atlas, TCGA) of whole-slide images from almost 11,000 patients. We successfully indexed and searched almost 30,000 high-resolution digitized slides constituting 16 terabytes of data comprised of 20 million 1000 × 1000 pixels image patches. The TCGA image database covers 25 anatomic sites and contains 32 cancer subtypes. High-performance storage and GPU power were employed for experimentation. The results were assessed with conservative "majority voting" to build consensus for subtype diagnosis through vertical search and demonstrated high accuracy values for both frozen section slides (e.g., bladder urothelial carcinoma 93%, kidney renal clear cell carcinoma 97%, and ovarian serous cystadenocarcinoma 99%) and permanent histopathology slides (e.g., prostate adenocarcinoma 98%, skin cutaneous melanoma 99%, and thymoma 100%). The key finding of this validation study was that computational consensus appears to be possible for rendering diagnoses if a sufficiently large number of searchable cases are available for each cancer subtype.

How I use molecular genetic tests to evaluate patients who have or may have myelodysplastic syndromes

Myelodysplastic syndromes (MDS) can be difficult to diagnose, especially when morphological changes in blood and marrow cells are minimal, myeloblast proportion is not increased, and the karyotype is normal. The discovery of >40 genes that are recurrently somatically mutated in MDS patients raised hope that molecular genetic testing for these mutations might help clarify the diagnosis in ambiguous cases where patients present with cytopenias and nondiagnostic marrow morphological findings. However, many older healthy individuals also harbor somatic mutations in leukemia-associated driver genes, especially in DNMT3A, TET2, and ASXL1, and detection of common aging-associated mutations in a cytopenic patient can cause diagnostic uncertainty. Despite this potential confounding factor, certain somatic mutation patterns when observed in cytopenic patients confer a high likelihood of disease progression and may allow a provisional diagnosis of MDS even if morphologic dysplasia and other diagnostic criteria are absent. A subset of acquired mutations also influences risk stratification of patients with an established MDS diagnosis and can inform treatment selection. Many unanswered questions remain about the implications of specific mutations, and clinicians also vary widely in their comfort with interpreting sequencing results. Here, I review the use of molecular genetic assays in patients with possible MDS or diagnosed MDS.