SU-E-J-186: CT Textures Can be Predictive for Tumor Shrinkage

Area of Science:

• Radiotherapy and Medical Imaging
• Oncology and Cancer Research
• Computational Biology and Bioinformatics

Background:

• Proton therapy is an advanced radiation technique for cancer treatment.
• Predicting treatment response in non-small cell lung cancer (NSCLC) is crucial for optimizing patient outcomes.
• Texture analysis of medical images offers potential biomarkers for treatment response.

Purpose of the Study:

• To investigate if computed tomography (CT) texture features of the gross tumor volume (GTV) in NSCLC can predict tumor shrinkage after proton therapy with concurrent chemotherapy.
• To explore the utility of quantitative 3D texture analysis for response assessment in NSCLC patients undergoing proton therapy.

Main Methods:

• 25 locally advanced NSCLC patients treated with 74 Gy proton therapy and chemotherapy were analyzed.
• Weekly 4D CT scans were used to track GTV changes via deformable image registration.
• 3D texture features were extracted from GTV CT images using MaZda software and analyzed using neural networks for classification of strong vs. weak responders.

Main Results:

• Six classification models were evaluated, achieving an average true positive rate (TPR) of 0.66 and an average classification accuracy of 64.8%.
• The models demonstrated moderate predictive capability for distinguishing between strong and weak responders based on GTV CT texture features.
• Average false positive rate (FPR) was 0.36, with an average one-sided p-value of 0.118.

Conclusions:

• CT texture analysis of NSCLC GTV shows promise as a predictive biomarker for tumor shrinkage following proton therapy.
• Further validation with larger patient cohorts, diverse texture metrics, and advanced analysis techniques is warranted.
• This approach could potentially aid in personalizing NSCLC treatment strategies.