Deep Learning Model for Identifying Organ at Risk Toxicity in Thorax Carcinomas Caused by Radiation Therapy

Abstract

To assist doctors in making individualized treatment decisions, a machine learning algorithm that can accurately predict toxicity could be used. Patients undergoing thoracic radiation treatment often have organ at risk (OAR) toxicity, which may have a major effect on their health. The use of radiation therapy to treat thorax cancer has been shown to be beneficial. Toxins associated with treatment, on the other hand, are a different story. Treatment-related toxicity is a significant issue because cancer spreads to healthy cells and organs in close proximity. The use of radionics in therapy control strategies has been studied in a number of research features extracted from computed tomography (CT), magnetic resonance imaging (MRI), or dosimetric features collected by 3D dose distributions (DD). Convolutional Neural Networks (CNN) are being used to investigate the toxicity of organs in danger. Image processing, treatment planning, treatment delivery, and post-therapy follow-up are all feasible with a deep learning model. As well as by optimising Using convolution filters during the training stage, a deep learning-based prediction model may discover and fine-tune variables for particular categorization issues. The potential of a deep learning-based estimate technique to reduce the radiation treatment hazard rate is a significant benefit. Furthermore, using convolution methods, the attributes may be obtained automatically (CNN). The objective of this study is to create and test a 3D CNN-based toxicity prediction model. CT images, radiation treatment DD, and contours will be used to extract low and large temporal spatial features utilising (3D) filters, especially in low and large temporal spatial characteristics. The comparative test will demonstrate that the planned model is capable of forecasting OAR toxicity properly. Research into more specific criteria for places that are closely connected to OAR Toxicity might improve the model even further.