Modelling a dense network for soft tissue prediction using pre-trained network

Medical image analysis is driven by the emergence of deep learning and the increased computationally available. Concurrently, leveraging pertinent deep features can substantially enhance the performance of trainable expert systems and intelligent systems, thereby reducing both diagnosis time and the laborious nature of the process. This study introduces a radiomics framework based on deep learning for aiding soft tissue sarcomas diagnosis. Magnetic Resonance (MR) images confirmed to have diagnosed Liposarcoma (LPS) and Leiomyosarcomas (LMS) histologically were sourced from the dataset. The investigation within this study aims to assess the importance and influence of deep learning accomplished by the medical domain. To achieve this goal, this work propose dense-layered convolutional neural network (DLCNN) with pre-trained VGG-16 model. Experimental outcomes demonstrated that the proposed fusion framework surpassed prevailing techniques. As an outcome, the accuracy was 99% with the proposed model highlighting complementary information. Encouraged by the outcomes, our framework has the potential to help radiologists in classifying LMS and LPS.