Automated Segmentation of Ablated Lesions Using Deep Convolutional Neural Networks: A Basis for Response Assessment Following Laser Interstitial Thermal Therapy

Abstract Background Laser interstitial thermal therapy (LITT) of intracranial tumors or radiation necrosis enables tissue diagnosis, cytoreduction, and rapid return to systemic therapies. Ablated tissue remains in situ, resulting in characteristic post-LITT edema associated with transient clinical worsening and complicating post-LITT response assessment. Methods All patients receiving LITT at a single center for tumors or radiation necrosis from 2015 to 2023 with ≥9 months of MRI follow-up were included. An nnU-Net segmentation model was trained to automatically segment contrast-enhancing lesion volume (CeLV) of LITT-treated lesions on T1-weighted images. Response assessment was performed using volumetric measurements. Results Three hundred and eighty four unique MRI exams of 61 LITT-treated lesions and 6 control cases of medically managed radiation necrosis were analyzed. Automated segmentation was accurate in 367/384 (95.6%) images. CeLV increased to a median of 68.3% (IQR 35.1–109.2%) from baseline at 1–3 months from LITT (P = 0.0012) and returned to baseline thereafter. Overall survival (OS) for LITT-treated patients was 39.1 (9.2–93.4) months. Lesion expansion above 40% from volumetric nadir or baseline was considered volumetric progression. Twenty-one of 56 (37.5%) patients experienced progression for a volumetric progression-free survival of 21.4 (6.0–93.4) months. Patients with volumetric progression had worse OS (17.3 vs 62.1 months, P = 0.0015). Conclusions Post-LITT CeLV expansion is quantifiable and resolves within 6 months of LITT. Development of response assessment criteria for LITT-treated lesions is feasible and should be considered for clinical trials. Automated lesion segmentation could speed the adoption of volumetric response criteria in clinical practice.