3206 - Innovative No-Reference Image Quality Assessment for MRI-Guided Radiotherapy: Automated Distortion Detection
Presenter(s)
Z. Wang1, S. Chen1, J. Dai2, Y. Tang3, G. Wu1, and J. Chen2; 1School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing, China, 2Department of Radiation Oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China, 3Department of Radiation Oncology, National Cancer Center/Cancer Hospital and Institute, Chinese Academy of Medical Sciences (CAMS) and Peking Union Medical College (PUMC), Beijing, China
Purpose/Objective(s):
To develop and evaluate a novel no-reference image quality assessment (NRIQA) method aimed at improving tumor tracking accuracy in MRI-guided radiotherapy (MRIgRT) by enhancing image quality and efficiently identifying image quality.Materials/Methods: A dataset of 106,000 MRI frames from 10 patients with liver metastasis, acquired using the MR-LINAC system from a precision radiation medicine company, was analyzed. The NRIQA method involves three components: (1) Image Preprocessing: Optimizing techniques to enhance diagnostic feature visibility. (2) Feature Extraction and Directional Analysis: Utilizing Mean Subtracted Contrast Normalized (MSCN) coefficients across four directions to capture texture and identify distortions. (3) Quality Index (QI) Calculation: Combining features through Asymmetric Generalized Gaussian Distribution (AGGD) parameter estimation and K-means clustering to provide a comprehensive quality measure. The Tracking-Learning-Detection (TLD) algorithm was used on both preprocessed and unprocessed images to assess tumor tracking performance. The QI’s effectiveness was compared with traditional metrics (Contrast-to-Noise Ratio [CNR], Tenengrad gradient, and entropy) through statistical analysis.
Results: Preprocessing significantly improved image quality, with tumor tracking precision from 78.6% to 94.9% and recall from 7.4% to 76.5%. The QI outperformed traditional metrics in image quality detection, showing improvements of 79.6 times over CNR, 6.5 times over Tenengrad gradient, and 1.7 times over entropy, highlighting its superior sensitivity and effectiveness in identifying image quality improvements.
Conclusion:
The NRIQA method demonstrated rapid detection of the effectiveness of image preprocessing and established a robust framework for evaluating image quality. By efficiently identifying and excluding suboptimal images, the method significantly improved tumor tracking efficiency. This study highlights the potential of NRIQA to enhance clinical workflows in MRI-guided radiotherapy by ensuring high-quality images for treatment planning and delivery.