1225 - Are Some "New" Lung Metastases Already Present at Baseline? An Artificial Intelligence-Based Re-evaluation in Patients Undergoing Lung SBRT for Oligometastases

08:10am - 08:15am ET

Room 156

Presenter(s)

Heejoo Ko, MD, BS - College of Medicine, The Catholic University of Korea, Seoul, SEO

H. Ko¹, J. Y. Lee², E. M. Dunne³, M. Liu³, K. J. Lee⁴, D. Y. Oh⁴, J. C. Park⁴, and J. Chang¹; ¹Department of Radiation Oncology, Yonsei University College of Medicine, Seoul, Korea, Republic of (South), ²Department of Radiation Oncology, Yonsei Cancer Center, Heavy Ion Therapy Research Institute, Yonsei University College of Medicine, Seoul, Korea, Republic of (South), ³BC Cancer Vancouver, Vancouver, BC, Canada, ⁴Monitor Corporation, Seoul, Korea, Republic of (South)

Purpose/Objective(s):

Accurate baseline lesion enumeration is essential for metastasis-directed therapy (MDT) in oligometastatic lung disease. We hypothesized that artificial intelligence–based computer-aided detection (AI-CAD) would reveal additional baseline pulmonary lesions not incorporated into MDT planning, and that these lesions would be associated with subsequent progression and repeat MDT.

Materials/Methods:

We retrospectively analyzed 118 patients who underwent lung stereotactic body radiation therapy (SBRT) for oligometastatic disease (153 treated lesions; median follow-up 3.7 years). The cohort included patients with diverse primary malignancies, most commonly colorectal (39.8%), hepatopancreatobiliary (21.2%), lung (14.4%), and genitourinary cancers (9.3%); 21.2% presented with oligoprogression. AI-CAD was applied to baseline simulation CT, baseline diagnostic CT, and follow-up CT scans. All detected pulmonary nodules were manually matched across timepoints for lesion-level tracking. Outcomes included detection of treated lesions, additional baseline nodules not selected for SBRT, subsequent radiographic enlargement, repeat MDT, and slice thickness–dependent detection.

Results:

AI-CAD detected 135 of 153 SBRT-treated lesions (88.2%); missed lesions were mainly subpleural or juxta-pleural lesions. At baseline, 76.7% of 580 detected nodules were not incorporated into MDT planning. During follow-up, 107 untreated nodules (24.0%) enlarged, corresponding to 43 patients (36.4%); 74 nodules (16.6%) met RECIST progression criteria. Repeat SBRT was administered in 41.5% of patients overall, with 22% of repeat treatments performed for previously untreated baseline lesions after a median of 260 days. Thin-slice (1 mm) diagnostic CT identified 367 additional nodules not visible on simulation CT, 21.3% of which subsequently enlarged.

Conclusion:

AI-CAD identified additional baseline pulmonary lesions, a subset of which later progressed and contributed to repeat MDT. Under-recognized baseline disease may partially account for subsequent intrapulmonary progression. AI-assisted comprehensive baseline pulmonary lesion assessment, particularly with thin-slice imaging, may improve staging accuracy and inform MDT decision-making.