Leukaemia and childhood cancers are becoming increasingly prevalent in India, highlighting the urgent need for advancements in early detection methodologies. This research paper explores the application of deep learning (DL) and its techniques in examining blood cells to aid in the diagnosis of blood cancers. DL algorithms possess the ability to accurately classify and segment cells in blood smears, thereby facilitating early-stage cancer detection. The accuracy of cell segmentation has significantly improved due to the utilization of advanced DL techniques such as generative adversarial networks (GANs) (Wang et al. in Pattern Recogn Lett 141:122–128, 2021), which effectively address challenges like overlapping nuclei and morphological changes. Additionally, convolutional neural networks (CNNs), enhanced through transfer learning methods (Alakus and Akkus in Comput Biol Med 143:104551, 2022), have exhibited remarkable precision in classifying blood cancer cells. The primary aim is to combine progress in deep learning with a Clinical Decision Support System (CDSS) that quickly converges to provide accurate diagnoses and improve patient outcomes is the primary objective. The incorporation of federated learning models with explainable AI (XAI) (Saraswat et al. in IEEE Access 10:84486–84517, 2022), which ensures privacy and transparency in data processing, has bolstered trust in DL applications. The shift in DL from segmentation to classification signifies a paradigm shift in how India tackles cancer detection.