Utilizing the Inception-ResNetV2 Pre-trained Model for Binary Classification of Leukemia Cells: An Advanced Approach to Hematological Diagnostics

Leukaemia, a form of hematologic malignancy, affects individuals across all age groups and is among the primary contributors to global mortality. Acute lymphoblastic leukaemia (ALL) is a prevalent form of leukaemia that mostly affects the bone-marrow in humans. The conventional methods employed for disease diagnoses, such as blood and bone marrow exams, are characterized by their sluggish and painful nature. To stop leukaemia from killing people, creating a computer-based automatic and strong classification system has become important. Algorithms developed using deep learning are widely utilized in contemporary medical practice for leukaemia treatment, namely in the identification and diagnosis of leukaemia in patients. This research presents an advanced approach in deep learning for the detection and categorization of leukaemia cells. To carry out this analysis, a Convolutional Neural Network model utilizing the Inception_Resnet50 architecture has been suggested. The Leukaemia Classification dataset, which contains 15114 images, is utilized to train and evaluate the model. The model has shown a very good accuracy of 95.75% in classifying the cells for ALL and Normal classes. This research can be implemented in the healthcare field for pre-diagnosis of Leukemia in blood cells.