Detecting Malignant Leukemia Cells Using Microscopic Blood Smear Images: A Deep Learning Approach

Leukemia is a form of blood cancer that develops when the human body’s bone marrow contains too many white cells. This medical condition affects adults and considered prevalent in children. Treatment for leukaemia determined by type extent to which has developed across body. It crucial diagnose early order provide adequate care cure patients. Researchers have been working on advanced diagnostics systems based Machine Learning (ML) approaches early. In this research, we employ deep learning (DL) convolutional neural network (CNN) hybridized two individual blocks CNN named CNN-1 CNN-2 detect acute lymphoblastic (ALL), myeloid (AML), multiple myeloma (MM). The proposed model detects malignant cells using microscopic smear images. We construct dataset about 4150 images from public directory. main challenges were background removal, ripping out un-essential components supplies, reduce noise blurriness minimal method image segmentation. To accomplish pre-processing segmentation, transform RGB color-space into greyscale 8-bit mode, enhancing contrast intensity adjustment adaptive histogram equalisation (AHE) method. increase structure sharpness multiplication binary with output enhanced next step, complement done get black colour nucleus colour. Thereafter, applied area operation closing remove noise. Finally, multiply final source regenerate space, resize [400, 400]. After applying all methods techniques, managed noiseless, non-blurred, sharped segmented lesion. are given as input CNNs. Two parallel CCN models trained, extract features. extracted features further combined Canonical Correlation Analysis (CCA) fusion more prominent used five classification algorithms, namely, SVM, Bagging ensemble, total boosts, RUSBoost, fine KNN, evaluate performance feature extraction algorithms. Among ensemble outperformed other algorithms achieving highest accuracy 97.04%.