Deep learning-based hybrid multi-task model for adrenocortical carcinoma segmentation and classification

Abstract

Adrenocortical Carcinoma (ACC) is a rare but highly lethal cancer that occurs in the adrenal cortex. Accurate diagnosis of ACC are vital in order to determine appropriate treatment strategies and predict patient outcomes. Hence, defining the stages of ACC is a crucial factor for both diagnosis and treatment planning and it is the key aspect that the researchers are still exploring. Our study proposes a novel deep learning-based hybrid Multi-Task model which performs both segmentation to find the exact cancer region and classification based on the cancer stages. Thus our model is resource efficient. In our research, several deep learning-based architectures have been used to segment and evaluate the ACC CT images. Moreover, we have explored how Convolutional Neural Network (CNN) classification models perform on the classification task. This process includes the exploration to find the model based on the Multi-Task learning model’s feature extraction perform on classification task.