Western University of Health Sciences, College of Osteopathic Medicine of the Pacific Pomona, CA
A. Lian1, T. Ketcherside2, C. Han2, A. Liu2, C. J. Ladbury2, and A. Amini2; 1College of Osteopathic Medicine, Western University, Pomona, CA, 2Department of Radiation Oncology, City of Hope National Medical Center, Duarte, CA
Purpose/Objective(s):There is interest in utilizing radiomics to monitor, and potentially act upon, treatment response in patients receiving radiation treatment. Historically, computed tomography (CT) imaging obtained for image-guided radiotherapy is of insufficient quality for use in radiomic analyses. A new linear accelerator that utilizes fan-beam kilovoltage CT (kVCT) has higher-quality imaging, creating opportunities for radiomic analysis. This study sought to evaluate the feasibility of using these scans to correlate radiomic feature changes with treatment response in patients undergoing definitive treatment for lung cancer.Materials/
Methods: Patients were scanned using the novel fan-beam kVCT system throughout treatment. CT scans used standardized acquisition parameters for all treatments to minimize variation. Images were acquired using “Body/Medium Dose/Slow Couch” parameters. The primary tumor gross tumor volume (GTV) was contoured by a radiation oncologist on each scan, which were anonymized and put in random order to prevent bias. CTs and structure files were imported into an Image Biomarker Standardization Initiative compliant radiomic software package (LifeX) to extract radiomic features. 40 radiomic features that were identified to be stable in normal tissues during treatment were extracted from each scan with the volume-of-interest defined as the GTV. Pearson correlation coefficients were calculated between these features and relative change in GTV volume (1 - {GTV volume/GTV volume at time of 1st fraction [fx]}). Coefficients greater than 0.4, 0.7, and 0.9 were considered moderately, strongly, and very strongly correlated, respectively. All distributions are expressed as median [IQR] unless otherwise stated. Results: 12 patients (9 male) with lung cancer (11 non-small cell lung cancer, 1 small cell lung cancer) treated with conventionally fractionated (60 Gy/30 fx [n=8] or 66 Gy/33 fx [n=3]) or hypofractionated (45 Gy/15 fx [n=1]) radiation between 9/2021 and 10/2023 were analyzed. Median age was 64 (range: 42-84) years. During treatment, median decrease in GTV volume was 42.3% (8.9-62.5%). One radiomic feature was very strongly correlated with GTV response: GLZLM_GLNU (0.96 [0.91-0.97]). Two features were strongly correlated: NGLDM_Coarseness (0.83 [0.72-0.92]) and NGLDM_Busyness (0.72 [0.46-0.79]). Four other features were moderately correlated: DISCRETIZED_HUpeakSphere0.5mL, DISCRETIZED_HUQ1, NGLDM_Contrast, and CONVENTIONAL_HUQ1. Conclusion: Three radiomic features were at least strongly correlated with primary GTV volume changes when evaluated on images obtained by a fan-beam kVCT in patients treated definitively for lung cancer. These findings demonstrate potential feasibility to use these images for radiomic analysis during treatment to predict individual patient response. Once sufficient follow-up is available, future work will focus on whether these features also correlate with overall treatment response and disease control.
