Designing an Explainable AI Framework for Leukemia Diagnosis: Integrating Customized CNN-SVM Models with Optimized Deep Learning

White blood cell appearance is one of the most important criteria for diagnosis for leukaemia. Since the conclusion of a diagnosis may be affected by the personal perspective of the person making the diagnosis, it is very advantageous to use a computerised methodology. To make this procedure easier, a doctor will look at blood smear slides under an electron microscope and make a diagnosis based on what they see. In terms of the deep learning frameworks, we are able to get measures like prospective accuracies. To get the most out of these frameworks, you need a big dataset with plenty of pictures. Due of its rarity, leukaemia databases often include photos of a smaller size. The suggested approach makes use of a tailored CNN integrated with a support vector machine. The classifier decides whether the cells are normal or infected. To verify the accuracy of this judgement, the widely used XAI framework LIME is used for interpretation. The present study makes use of the widely used and freely accessible ALL-IDB dataset. The suggested technique follows the beginning and follows by a presentation of relevant work. The findings, discussions, and conclusions are all included in the following section. It goes on to say that with a 96% accuracy rate, the suggested CNN-SVM method is much superior than GradCAM, Shaply, and LIME. The CNN-SVM model outperformed the others in terms of accuracy, whereas GradCAM outperformed them across the board. The explanations provided by Shaply and LIME were also rather accurate.