Development of Automated Python-based Software for RBC Image Analysis: An Open-source Application Towards Early Diagnosis of Leukemia
Ritobrita Sen, Amitava Choudhury, Ishaan Majumder, Sumita Biswas, Ishita Chatterjee, Durjoy MajumderBackground/Introduction:
Early detection of hematological malignancies improves therapeutic outcomes, and digital pathology aids this, thus supporting clinical decision-making. A few image- analysis tools exist, but most are available only on proprietary platforms or require computational programming expertise, limiting their widespread adoption in routine diagnostics. To address this gap, we developed a fully automated, Mathematical Morphology (MM)-based image-analysis framework in Python, an open-source platform, to identify Red Blood Cell (RBC) morphological alterations in Acute Lymphoblastic Leukemia (ALL).
Methods:
MM algorithms were used for segmentation, binary and grayscale operations (intensity levels range 0-255) to enhance structural characteristics following morphological operations (dilation, erosion, opening, closing) for feature extraction from RBC images in a quantitative manner. RBC images of peripheral blood of 10 ALL cases (542 RBCs) and 5 healthy individuals (104 RBCs) were analyzed. Four features-area, perimeter, convexity, and solidity-were computed, and group differences were assessed using two-tailed t-tests with significance at p < 0.05.
Results:
Leukemic RBCs exhibited significantly reduced areas and perimeter compared with normal controls (p < 0.05), indicating size reduction with subtle boundary changes. Convexity and solidity did not differ significantly between groups, suggesting preservation of overall shape integrity despite reductions in cell dimensions.
Discussion:
This study presents an MM-based Python platform that detects morphological differences in RBCs associated with ALL. By matching the performance of earlier MATLAB-based methods while eliminating proprietary barriers, the proposed framework offers a cost-effective, accessible, and scalable tool for digital pathology.
Conclusion:
The developed software demonstrated its utility in early screening, underscoring its relevance in preventive oncology and positioning it to advance point-of-care diagnostic capabilities.