A Novel Automated Approach for Leukemia Diagnosis using Digital Image Processing and Machine Learning

This study introduces a novel automated strategy for diagnosing leukemia through the application of digital image processing techniques. To isolate individual cells in ultra-high resolution blood stain images, the proposed method utilizes state-of-the-art image segmentation. The subsequent phase in the process of accurately identifying leukemic cells is feature extraction, which is designed to emphasize critical cellular characteristics. Then, a machine learning classifier is employed to accurately differentiate between normal and malignant cells. The method’s stated objective is to provide a non-invasive, costeffective, and reliable alternative to current remedies, while also streamlining the diagnostic process. Compared to a reference dataset, this technique’s exceptional sensitivity, specificity, and overall accuracy may be advantageous for clinical leukemia surveillance and early detections. The proposed procedure achieved a precision of $93.2 \%$ and a recall of $96.8 \%$, resulting in a $95.4 \%$ accuracy.