Deep Learning Based Radiomics for Prognostic Prediction in Lung Cancer Patients

Abstract

Recently, radiomics was studied positively in the field of radiotherapy. There are two methods for radiomics analysis: One is handcrafted-based radiomics, the other is deep learning-based radiomics. Handcrafted-based radiomics workflow is typically based on extracting human-defined features from as segmented region. On the other hand, deep learning-based radiomics dose not need any prior knowledge and features. Another benefit of deep learning based radiomics is that the input data to the deep networks to extract radiomics features is the raw image without segmenting the region of interest. For this reason, we donʼt need to prepare the segmentation. In this review paper, we focused on the radiomics for prognostic prediction of lung cancer patients treated with radiotherapy, and explained the two different radiomics analysis with our experience.