AN EFFECTIVE VISION TRANSFORMER-AIDED INTELLIGENT EFFICIENTNET B7 WITH ADAPTIVE TUMOR SEGMENTATION FOR BREAST CANCER DIAGNOSIS USING THERMOGRAPHY IMAGES

Breast cancer is a very severe disease in the medical sector. This is the second deadliest disorder among women after lung tumors. Women troubled with breast tumors are susceptible to grave health-related problems with higher death rates. Hence, it has become an important concern for healthcare specialists. However, timely and early intervention utilizing advanced projection to identify and diagnose certain difficulties can improve the survival chances of the tumor patients. The breast tumor is the most common tumor, which is the reason for more than a thousand women’s deaths every year. The timely recognition of the breast tumor minimizes the rate of death. Thus, breast tumor screening is significant in the initial stages. However, not many improving regions can afford the detection and screening strategies because of their cost. Therefore, an efficient and low-cost way of identifying the breast tumor is conducted utilizing thermography that, unlike other approaches, can be employed in women of different age categories. In this work, a novel deep learning-assisted breast cancer detection mechanism is offered. In the beginning, the images are accumulated from the standard dataset. The garnered images are rendered to the segmentation process. Here, the thermogram images are segmented by the “Adaptive Fuzzy C Mean clustering (AFCM)”. Here, the parameter in the AFCM is tuned by the modified prey–predator algorithm (MPPA). Further, the segmented thermogram images are forwarded to the diagnosis phase. Here, the Vision Transformer-based Attention and Multiscale Dilated EfficientNet B7 (ViT-AMDNet) is introduced to perform the diagnosis process. At last, the results achieved from the developed model are compared with the existing breast cancer identification techniques to determine the efficacy of the developed model.