ENHANCING MORPHOLOGICAL ANALYSIS OF PERIPHERAL BLOOD CELLS IN CHRONIC LYMPHOCYTIC LEUKEMIA WITH AN ARTIFICIAL INTELLIGENCE‐BASED TOOL

Abstract

Background: Real-time monitoring of chronic lymphocytic leukemia (CLL) is crucial for effective patient management. Peripheral blood (PB) is the preferred source for obtaining CLL cells due to its ease of access and lower cost. However, traditional methods of evaluating PB films have several drawbacks, such as lack of automation, reliance on personal experience, and low repeatability and reproducibility. To address these limitations, we developed an artificial intelligence-based tool that objectively evaluates morphologic features in the blood cells of CLL patients from a clinical perspective. Methods: We conducted a retrospective study on 284 patients newly diagnosed with CLL at Jiangsu Province Hospital from December 1, 2013 to December 31, 2020. To train and test our machine-learning algorithm, we randomly split the cohort into a training set (75%) and a testing set (25%). We obtained high-quality whole-slide images from blood films using the Bionovation CSFA-80 scanner, and developed an automated algorithm using a deep convolutional neural network (CNN) to precisely identify regions of interest. Additionally, we used the well-established Visual Geometry Group-16 CNN as the encoder to segment cells and extract morphological features. This tool enabled us to extract precise morphological features of all lymphocytes for subsequent analysis. Results: Our study's lymphocyte identification had a recall of 0.96 and an F1 score of 0.97, making it suitable for future applications. Cluster analysis identified three clear, morphological groups of lymphocytes that reflect distinct stages of disease development to some extent. To investigate the longitudinal evolution of lymphocyte, we extracted cellular morphology parameters at various time points from the same patient. These time points included the initial diagnosis period without treatment indications, the pretreatment period with indications beginning to appear, and the diagnosed phase of large cell transformation. The results showed some similar trends to those observed in the aforementioned cluster analysis. Correlation analysis further supports the prognostic potential of cell morphology-based parameters. The research was funded by: National Natural Science Foundation of China [82100211]; Beijing Xisike Clinical Oncology Research Foundation [Y-Roche2019/2-0090]; Science Foundation Project of Ili & Jiangsu Joint Institute of Health [yl2021ms04]; and Science and Technology Development Fund Project of Pukou branch of Jiangsu People's hospital [KJ2021-4]. Keywords: diagnostic and prognostic biomarkers, Imaging and Early Detection - Other, chronic lymphocytic leukemia (CLL) No conflicts of interests pertinent to the abstract.