Iranian Congress of Radiology

Iranian Congress of Radiology

EVALUATING THE EFFECTIVENESS OF ARTIFICIAL INTELLIGENCE FOR ORGAN SEGMENTATION IN IMAGE-GUIDED RADIOTHERAPY: A SYSTEMATIC REVIEW

Authors
Amin Banaei
Samaneh Hassanpour
Ali Vafadar
Department of Medical Physics, Faculty of Medical Sciences, Tarbiat Modares University, Tehran, Iran
10.22034/icrj.2022.173692
Abstract
Background:
Image-guided radiotherapy (IGRT) provides high accurate treatment for cancer patients, and better contouring and preservation of organs at risk (Oars) by tracking tumors motions and deformations using CT and MRI images obtained during treatment(1). Manual contouring of OAR and target volumes are complicated and time-consuming (2). Nowadays, artificial intelligence (AI) is considered a help, which can be used to segment the structures in CT and MRI images for treatment planning, patient setup verification, and evaluating tumor deformations (3). This study aims to express and conclude the effectiveness of AI in IGRT.
Methods:
The literature search was conducted in pubmed, sciencedirect, and google scholar using terms “MRI”, “Radiotherapy”, ” IGRT”, “Artificial Intelligence”, “Machine Learning” and “Auto-Segmentation”. The most relevant articles were selected and reviewed to evaluate the effectiveness of AI in IGRT.
Results:
Several studies showed that the use of AI in IGRT can resulted to significant time reduction compared to manual contouring in head and neck (4, 5), prostate (6), lung (7) and rectal(8) cancers. Furthermore, several studies reported high delineation accuracy for AI on CT images (9-11). However, some studies indicated that the accuracy of the structures delineation using AI, especially for CT images with insufficient clarity and contrast, resulted in the wrong delineation and subsequently over or under-dosage of patients (12, 13). It is suggested that MRI images with superior soft-tissue contrast can help to obtain better structure delineation (14). However, commercial CT scanners combined with radiotherapy machines are more accessible than MRI systems in the clinic.
Conclusion:
As manual organ delineation takes significant time and effort, there will be an increase in using AI for organ delineation in the future. MRI images for AI organ delineation can result in higher accuracy contouring than CT images; however, output contoured organs must be reviewed and modified with humans before any radiation dose calculations.
Iranian Congress of Radiology
Volume 37, Issue 3
Summer 2022
Pages 47-47

  • Receive Date 29 June 2023