Hematologic cancer diagnosis and classification using machine and deep learning: State-of-the-art techniques and emerging research directives.

Hematology is the study of diagnosis and treatment options for blood diseases, including cancer. Cancer is considered one of the deadliest diseases across all age categories. Diagnosing such a deadly disease at the initial stage is essential to cure the disease. Hematologists and pathologists rely on microscopic evaluation of blood or bone marrow smear images to diagnose blood-related ailments. The abundance of overlapping cells, cells of varying densities among platelets, non-illumination levels, and the amount of red and white blood cells make it more difficult to diagnose illness using blood cell images. Pathologists are required to put more effort into the traditional, time-consuming system. Nowadays, it becomes possible with machine learning and deep learning techniques, to automate the diagnostic processes, categorize microscopic blood cells, and improve the accuracy of the procedure and its speed as the models developed using these methods may guide an assisting tool. In this article, we have acquired, analyzed, scrutinized, and finally selected around 57 research papers from various machine learning and deep learning methodologies that have been employed in the diagnosis of leukemia and its classification over the past 20 years, which have been published between the years 2003 and 2023 by PubMed, IEEE, Science Direct, Google Scholar and other pertinent sources. Our primary emphasis is on evaluating the advantages and limitations of analogous research endeavors to provide a concise and valuable research directive that can be of significant utility to fellow researchers in the field.

Fifth Edition of the World Health Organization Classification of Tumors of the Hematopoietic and Lymphoid Tissues: Acute Lymphoblastic Leukemias, Mixed-Phenotype Acute Leukemias, Myeloid/Lymphoid Neoplasms With Eosinophilia, Dendritic/Histiocytic Neoplasms, and Genetic Tumor Syndromes.

This manuscript represents a review of lymphoblastic leukemia/lymphoma (acute lymphoblastic leukemia/lymphoblastic lymphoma), acute leukemias of ambiguous lineage, mixed-phenotype acute leukemias, myeloid/lymphoid neoplasms with eosinophilia and defining gene rearrangements, histiocytic and dendritic neoplasms, and genetic tumor syndromes of the 5th edition of the World Health Organization Classification of Tumors of the Hematopoietic and Lymphoid Tissues. The diagnostic, clinicopathologic, cytogenetic, and molecular genetic features are discussed. The differences in comparison to the 4th revised edition of the World Health Organization classification of hematolymphoid neoplasms are highlighted.

Blood cancer prediction using leukemia microarray gene data and hybrid logistic vector trees model.

Blood cancer has been a growing concern during the last decade and requires early diagnosis to start proper treatment. The diagnosis process is costly and time-consuming involving medical experts and several tests. Thus, an automatic diagnosis system for its accurate prediction is of significant importance. Diagnosis of blood cancer using leukemia microarray gene data and machine learning approach has become an important medical research today. Despite research efforts, desired accuracy and efficiency necessitate further enhancements. This study proposes an approach for blood cancer disease prediction using the supervised machine learning approach. For the current study, the leukemia microarray gene dataset containing 22,283 genes, is used. ADASYN resampling and Chi-squared (Chi2) features selection techniques are used to resolve imbalanced and high-dimensional dataset problems. ADASYN generates artificial data to make the dataset balanced for each target class, and Chi2 selects the best features out of 22,283 to train learning models. For classification, a hybrid logistics vector trees classifier (LVTrees) is proposed which utilizes logistic regression, support vector classifier, and extra tree classifier. Besides extensive experiments on the datasets, performance comparison with the state-of-the-art methods has been made for determining the significance of the proposed approach. LVTrees outperform all other models with ADASYN and Chi2 techniques with a significant 100% accuracy. Further, a statistical significance T-test is also performed to show the efficacy of the proposed approach. Results using k-fold cross-validation prove the supremacy of the proposed model.

Hematological malignancies in the Northwest Ethiopia.

BACKGROUND: The effect of malignant diseases is increasing globally, particularly in developing countries as shown by recent cancer statistics from the world health organization reports. It is anticipated that with an increase in life expectancy consequent upon the improved standard of living and increasing urbanization, the burden of hematological malignancies in sub-Saharan Africa particularly in Ethiopia is likely to increase recently. Therefore, this study was aimed to determine the incidence and trend of hematological malignancy in Northwest Ethiopia. METHODS: A facility-based retrospective study was conducted from 2015 to 2019 at the University of Gondar and Bahir-Dar Felegehiwot comprehensive specialized hospitals. Hematological malignancy data were collected by using a data collection sheet that was consisted of patients' socio-demography, clinical, and laboratory data. Then, data were entered into Epi-info 3.5.1 and exported to SPSS version 20 for analysis. Skewness and kurtosis were used to check data distribution. Descriptive statistics were summarized as percentages, means, and standard deviations of background variables, and the trend were analyzed. RESULTS: In this study, a total of 1,342 study participants were included. The mean age of study participants was 41.49 ± 16.3 years with a range of 1 to 92 years. About 58.3%, 52.2%, and 80% of the cases were observed among males, 18-45 age group, and urban residences, respectively. Of the total cases, 92.9% and 7.1% were lymphoma and leukemia, respectively. On the other hand, from lymphoma cases, 72.3% and 27.7% were HL and NHL, respectively while from leukemic cases, 61.1%, 23.2, 6.3%, 4.2%, and 5.3% were CLL, ALL, CML, AML, and other HM types, respectively. In this study, there was no trend. CONCLUSION: We concluded that lymphoma was the dominant type of hematological malignancy observed in northwest Ethiopia. The study indicated that the majority of cases were observed among male, urban residents, and adult populations aged 18-45 years. Therefore, special focus should be given to the highly affected population. Further, a prospective cohort study should be conducted for a better understanding of the prevalence and associated factors to it.

Is race important in genomic classification of hematological neoplasms?

In recent years, genome-based classifications for hematological neoplasms have been proposed successively and proved to be more accurate than histologic classifications. However, some previous studies have reported the racial differences of genetic landscape in persons with hematological neoplasms including myelodysplastic syndromes (MDS), which may cause a genomic classification based on a particular ethnic group does not operate in other races. To determine whether race plays an important role in the genomic-based classification, we validated a newly proposed genomic classification of MDS (J Clin Oncol.2021; JCO2001659), which was based on a large European database, in Chinese patients from our center. Our results showed significant differences between Chinese and European patients including proportion of each group to overall cohort when applying this novel genomic classification. Our data indicate that a genomic classification of hematological neoplasms probably should be revised according to specific genetic features in different races.

Mastocytosis, MCAS, and Related Disorders-Diagnosis, Classification, and Therapy.

Mastocytosis is a heterogeneous group of hematologic neoplasms defined by an accumulation of neoplastic mast cells (MC) in the skin, bone marrow, and other visceral organs [...].

Patterns of Human Leukocyte Antigen Class I and Class II Associations and Cancer.

Human leukocyte antigen (HLA) gene variation is associated with risk of cancers, particularly those with infectious etiology or hematopoietic origin, given its role in immune presentation. Previous studies focused primarily on HLA allele/haplotype-specific associations. To answer whether associations are driven by HLA class I (essential for T-cell cytotoxicity) or class II (important for T-cell helper responses) genes, we analyzed GWAS from 24 case-control studies and consortia comprising 27 cancers (totaling >71,000 individuals). Associations for most cancers with infectious etiology or of hematopoietic origin were driven by multiple HLA regions, suggesting that both cytotoxic and helper T-cell responses are important. Notable exceptions were observed for nasopharyngeal carcinoma, an EBV-associated cancer, and CLL/SLL forms of non-Hodgkin lymphomas; these cancers were associated with HLA class I region only and HLA class II region only, implying the importance of cytotoxic T-cell responses for the former and CD4+ T-cell helper responses for the latter. Our findings suggest that increased understanding of the pattern of HLA associations for individual cancers could lead to better insights into specific mechanisms involved in cancer pathogenesis. SIGNIFICANCE: GWAS of >71,000 individuals across 27 cancer types suggest that patterns of HLA Class I and Class II associations may provide etiologic insights for cancer.

Predictors of intensive care unit admission in patients with hematologic malignancy.

Limited data exist on predictors of intensive care unit (ICU) admission in patients with hematologic malignancy. The objective of this study was to identify predictors of ICU admission in hospitalized patients with hematologic malignancies. A retrospective cohort study was conducted on 820 consecutive admissions of patients with a malignant hematology diagnosis at our institution between March 2009 and December 2015. Backward stepwise selection procedure was conducted for multivariable logistic regression analyses. 820 patients were included, of whom 179 (22%) were admitted to the ICU. Types of hematologic cancers included 71% (N = 578) lymphoid cancer, 18% (N = 151) myeloid cancer, and 10% (N = 80) plasma cell neoplasms. 14% (N = 111) of patients had acute leukemia. Six predictors of admission to ICU were found in multivariable analysis, including disease-related (acute leukemia, curative intent chemotherapy), laboratory-related (platelet count < 50 × 109/L, albumin below normal, LDH above normal at time of admission), and physician-related factors (having advanced directives discussion) (p < 0.0001). A significant proportion of patients with hematologic malignancies admitted to hospital are admitted to ICU. Utilizing the identified predictors of ICU admission may help guide timely informed goals of care discussions with patients before clinical deterioration occurs.

Molecular tumor classification using DNA methylome analysis.

Tumor classifiers based on molecular patterns promise to define and reliably classify tumor entities. The high tissue- and cell type-specificity of DNA methylation, as well as its high stability, makes DNA methylation an ideal choice for the development of tumor classifiers. Herein, we review existing tumor classifiers using DNA methylome analysis and will provide an overview on their emerging impact on cancer classification, the detection of novel cancer subentities and patient stratification with a focus on brain tumors, sarcomas and hematopoietic malignancies. Furthermore, we provide an outlook on the enormous potential of DNA methylome analysis to complement classical histopathological and genetic diagnostics, including the emerging field of epigenomic analysis in liquid biopsies.

Contributing factors to the plasma albumin level at diagnosis of hematological malignancy.

OBJECTIVES: Many factors contribute to the plasma albumin (PA) level. We aimed to quantify different factors' relative contribution to the PA level when diagnosing hematological malignancy (HM). METHODS: The study was a population-based registry study including patients with HM in a Danish region. We applied multivariate linear regression analyses with C-reactive protein (CRP), WHO performance score (WHO-PS), age, sex, comorbidity, and HM type as exposures and the PA level on the day of the HM diagnosis (DX) as the outcome. The relative contribution of each exposure was determined as a percentage of the models' coefficient of determination (R2). RESULTS: In total, 2528 patients with HM had PA measured on DX. In the model comprising all exposures, CRP contributed with 65.8% to the R2 of 0.389 whereas 3 variables (CRP, WHO-PS, HM type) together contributed with 96.1%. When CRP was excluded from the model, R2 declined to 0.215 and the WHO-PS contributed with 96%. Other models, including separate analyses for each HM type, corroborated these results, except in myeloma patients where WHO-PS contributed with 61.1% to the R2 of 0.234. CONCLUSION: The inflammation biomarker CRP was the main predictor of the PA level on DX. The WHO-PS also contributed to the PA level on DX whereas the remaining factors (HM type, age, sex, and comorbidity) were of much less importance.

PDE4 subtypes in cancer.

Cyclic nucleotide phosphodiesterases (PDE) break down cyclic nucleotides such as cAMP and cGMP, reducing the signaling of these important intracellular second messengers. Several unique families of phosphodiesterases exist, and certain families are clinically important modulators of vasodilation. In the current work, we have summarized the body of literature that describes an emerging role for the PDE4 subfamily of phosphodiesterases in malignancy. We have systematically investigated PDE4A, PDE4B, PDE4C, and PDE4D isoforms and found evidence associating them with several cancer types including hematologic malignancies and lung cancers, among others. In this review, we compare the evidence examining the functional role of each PDE4 subtype across malignancies, looking for common signaling themes, signaling pathways, and establishing the case for PDE4 subtypes as a potential therapeutic target for cancer treatment.

Genetic predisposition in acute leukaemia.

A small but important proportion of patients with myelodysplasia (MDS) and acute leukaemia (AL) have underlying germline mutations in leukaemia susceptibility genes. The majority of these variants predispose to myeloid neoplasms with a smaller number associated with acute lymphoblastic leukaemia (ALL). The 2016 revision of the WHO classification of tumours of haematopoietic and lymphoid tissues has defined a number of myeloid neoplasms with germline predisposition (Blood, 127, 2016, 2391) alerting clinicians to the importance of this underlying diagnosis. Advances in genetic technology and access to testing will undoubtably result in increased numbers of patients and families with leukaemia predisposition syndromes being identified. Here we summarize the salient biology and genetic and clinical features of a number of these conditions including some more recently described genetic variants.

Targeted sequencing in DLBCL, molecular subtypes, and outcomes: a Haematological Malignancy Research Network report.

Based on the profile of genetic alterations occurring in tumor samples from selected diffuse large B-cell lymphoma (DLBCL) patients, 2 recent whole-exome sequencing studies proposed partially overlapping classification systems. Using clustering techniques applied to targeted sequencing data derived from a large unselected population-based patient cohort with full clinical follow-up (n = 928), we investigated whether molecular subtypes can be robustly identified using methods potentially applicable in routine clinical practice. DNA extracted from DLBCL tumors diagnosed in patients residing in a catchment population of ∼4 million (14 centers) were sequenced with a targeted 293-gene hematological-malignancy panel. Bernoulli mixture-model clustering was applied and the resulting subtypes analyzed in relation to their clinical characteristics and outcomes. Five molecular subtypes were resolved, termed MYD88, BCL2, SOCS1/SGK1, TET2/SGK1, and NOTCH2, along with an unclassified group. The subtypes characterized by genetic alterations of BCL2, NOTCH2, and MYD88 recapitulated recent studies showing good, intermediate, and poor prognosis, respectively. The SOCS1/SGK1 subtype showed biological overlap with primary mediastinal B-cell lymphoma and conferred excellent prognosis. Although not identified as a distinct cluster, NOTCH1 mutation was associated with poor prognosis. The impact of TP53 mutation varied with genomic subtypes, conferring no effect in the NOTCH2 subtype and poor prognosis in the MYD88 subtype. Our findings confirm the existence of molecular subtypes of DLBCL, providing evidence that genomic tests have prognostic significance in non-selected DLBCL patients. The identification of both good and poor risk subtypes in patients treated with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) clearly show the clinical value of the approach, confirming the need for a consensus classification.

Hematolymphoid Neoplasms Rarely Mimic Undifferentiated Pleomorphic Sarcoma of Soft Tissue.

CONTEXT.­: Undifferentiated pleomorphic sarcoma (UPS) of soft tissue is defined as a sarcoma with no recognizable line of differentiation. During the past few decades, advances in ancillary studies and review of prior UPS diagnoses have narrowed the category of UPS by excluding more-specific malignancies. However, few of those studies have specifically targeted pleomorphic hematolymphoid neoplasms. OBJECTIVE.­: To determine what fraction of UPS cases are misclassified pleomorphic hematolymphoid neoplasms, such as anaplastic large cell lymphoma, diffuse large B-cell lymphoma, histiocytic sarcoma (HS), myeloid sarcoma, and follicular dendritic cell sarcoma. DESIGN.­: Sixty-one UPS cases were screened by tissue microarray and an immunostain panel with subsequent analysis on whole block sections for suspicious cases. RESULTS.­: Five of 61 tumors (8%) were suggestive of HS based on the screening panel and were further evaluated with additional immunostains (PU.1, CD45, CD163) using whole sections. The 5 candidate HS cases were only focally positive for at most one stain with most staining in smaller, less-pleomorphic cells. Ultimately, no UPS met criteria for anaplastic large cell lymphoma, diffuse large B-cell lymphoma, myeloid sarcoma, follicular dendritic cell sarcoma, or HS. CONCLUSIONS.­: Our results suggest that a UPS of somatic soft tissue is unlikely to represent a misclassified hematopoietic malignancy. Exclusion of HS is most challenging, but immunostaining for PU.1, a nuclear transcription factor, may be easier to interpret in this context.

Hematologic Malignancies.

Radiation therapy plays a critical role in the management of a wide range of hematologic malignancies. The optimal radiation dose and target volume, and safe and effective ways of integrating radiation with systemic agents, vary depending on the histologic subtypes, stage at presentation, patient performance status, response to systemic therapy if given, treatment intent, and patient preferences. Limiting doses to surrounding organs without sacrificing disease control is of paramount importance. Reducing radiation doses and treatment volume in selected cases, and the use of advanced radiotherapy technology, can improve the therapeutic ratio of patients receiving radiation therapy for hematologic malignancies.

Refined Disease Risk Index for Hematological Malignancies, Including Rare Disorders, After Allogeneic Stem Cell Transplantation.

OBJECTIVE: The refined disease risk index (R-DRI) is a well-designed prognostic parameter that is based on only the disease type and status and is used for stratifying patients undergoing allogeneic hematopoietic stem cell transplantation (allo HSCT) into 4 risk groups. However, the application of the R-DRI for rare diseases has remained unclear. METHODS: We evaluated 135 patients who underwent allo HSCT for hematological malignancies including rare diseases, such as acute leukemia of ambiguous lineage, acute T-cell leukemia/lymphoma, extranodal natural killer T-cell lymphoma, and lymphoblastic lymphoma, at our institute. RESULTS: According to the R-DRI, overall survival (OS) and progression-free survival at 2 years for patients with the low, intermediate, high, and very high groups were 66.7% and 66.7%, 60.8% and 56.0%, 27.1% and 23.7%, and 5.9% and 5.1%, respectively (P < .0001 and P < .0001, respectively). OS showed no significant difference between B-cell non-Hodgkin lymphoma (B-NHL) and T-cell non-Hodgkin lymphoma (T-NHL) (P = .71). Moreover, OS at 1 year was 80%, 14.3%, 60%, and 0% for the intermediate risk group, the very high-risk group of B-NHL, the intermediate risk group, and the high-risk group of T-NHL, respectively (P = .035). CONCLUSION: We showed the applicability of the R-DRI for hematological malignancies, including rare disorders. However, we suggest that T-NHL patients may be better to be assigned between the nodal group and the extranodal group in the R-DRI.

Dark-matter matters: Discriminating subtle blood cancers using the darkest DNA.

The confluence of deep sequencing and powerful machine learning is providing an unprecedented peek at the darkest of the dark genomic matter, the non-coding genomic regions lacking any functional annotation. While deep sequencing uncovers rare tumor variants, the heterogeneity of the disease confounds the best of machine learning (ML) algorithms. Here we set out to answer if the dark-matter of the genome encompass signals that can distinguish the fine subtypes of disease that are otherwise genomically indistinguishable. We introduce a novel stochastic regularization, ReVeaL, that empowers ML to discriminate subtle cancer subtypes even from the same 'cell of origin'. Analogous to heritability, implicitly defined on whole genome, we use predictability (F1 score) definable on portions of the genome. In an effort to distinguish cancer subtypes using dark-matter DNA, we applied ReVeaL to a new WGS dataset from 727 patient samples with seven forms of hematological cancers and assessed the predictivity over several genomic regions including genic, non-dark, non-coding, non-genic, and dark. ReVeaL enabled improved discrimination of cancer subtypes for all segments of the genome. The non-genic, non-coding and dark-matter had the highest F1 scores, with dark-matter having the highest level of predictability. Based on ReVeaL's predictability of different genomic regions, dark-matter contains enough signal to significantly discriminate fine subtypes of disease. Hence, the agglomeration of rare variants, even in the hitherto unannotated and ill-understood regions of the genome, may play a substantial role in the disease etiology and deserve much more attention.