TICU-net: MediCareVision Framework for Disease Diagnosis using Cascaded U-net and Transformers Deep Leaning Models

The biomedical research task of diagnosing Hematological malignancies, like Leukemia and Myeloma, is a formidable challenge because it is machine automated. It is essential to the immune response against pathogens and foreign substances for white blood cells (WBCs). Early identification of blood cancer will result in fewer deaths due to its ability to identify disease, which encourages treatment and possible cure. Additionally, many of the investigations examine a restricted variety of WBC kinds consequently raising precision. The approach showcased is novel in proposing comprehensive pre-processing and data augmentation pipelines to generate a larger sized feature space for better results. The aim of this research is the building and creation of an automated system to detect white blood cell cancer illness using the Transformer Integrated Cascaded U-Net (TICU-net). It includes Myeloma and Leukemia (AML and ALL) at early stages. The input here is microscopic blood pictures. Further, we transformed the image into the right format (YCbCr) for segmentation. To segment, we used Otsu Adaptive Thresholding and for clustering we apply K-Means approach. The Gray Level Cooccurrence Matrix (GLCM) were used to extract features and then classified with the recommended deep learning model of Transformer Integrated Cascaded U-Net (TICU-nets). We compare the simulation experiments of few deep learning models to the TICU-net and our suggested system shows an accuracy of 98.95%.