3431 - MRI-Simulator and MR-Linac in Esophageal and Gastric Cancers: An Imaging-Pathology Correlation Study
Presenter(s)
S. C. Fong1,2, E. Lau3, N. Tebbutt4, R. Khor5, K. Brown5, F. Wightman5, B. Harris1, S. Fisher1, D. Williams6, S. Uribe7, and S. P. Ng7; 1Department of Radiation Oncology, Austin Health, Melbourne, VIC, Australia, 2Department of Radiation Oncology, Peter MacCallum Cancer Centre, Melbourne, VIC, Australia, 3Department of Radiology, Austin Health, Melbourne, VIC, Australia, 4Olivia Newton-John Cancer Wellness & Research Centre, Melbourne, Australia, 5Department of Radiation Oncology, ONJ Centre, Austin Hospital, Melbourne, VIC, Australia, 6Department of Pathology, ONJ Centre, Austin Hospital, Melbourne, Australia, 7Monash University, Melbourne, Australia
Purpose/Objective(s): To prospectively evaluate the relationship between MRI and histopathological tumor size measurements in patients undergoing surgical resection using both MR-simulator (1.5T) and MR-Linac (1.5T Unity) imaging, and to explore whether tumor characteristics influence measurement variations.
Materials/Methods: In this observational cohort study, 15 patients undergoing primary surgery without neoadjuvant therapy were enrolled. All patients underwent pre-operative MRI within 2 weeks of surgery on both a 1.5T technology company MR-simulator and a 1.5T MR-Linac from a precision radiation medicine company. T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI) sequences were obtained. Tumor characteristics (site, histology, grade, lymphovascular invasion) and pathological measurements were recorded. MRI measurements were performed by a radiologist blinded to histopathology results. Bland-Altman analysis and Pearson’s correlation were used to assess agreement between measurements.
Results: The cohort included 10 males and 5 females with a median age 75 years (range 51-87). Primary sites were gastric (n=9), distal/GOJ (n=5), and mid esophagus (n=1) with predominantly adenocarcinoma (14/15, 93.3%). Mean tumor measurement on MRI was smaller by 4.8 mm compared to histopathology (mean MRI: 37.2 mm, IQR: 23-50 mm; mean pathologic: 32.4 mm, IQR: 20-43 mm). Pearson's correlation was strong (r=0.90, p<0.001). Among MRI-visible tumors (n=10), measurement differences varied by size: tumors <30 mm (n=5) showed a mean overestimation of 1.8 mm (IQR: -3 to 7 mm), while tumors =30 mm (n=5) showed a larger mean overestimation of 7.8 mm (IQR: 2 to 7 mm).
Of 10 MRI-visible tumors, MRI exclusively detected disease in 7 cases where either CT or PET was negative, with DWI aiding detection in 3 cases. MRI detected two tumors that were negative on both CT and PET. In 5 cases without MRI-visible disease (mean histopathological size 11.0 mm, IQR: 3-20 mm), four were negative on all imaging modalities. One post-endoscopic submucosal dissection case showed false-positive CT/PET findings while MRI correctly demonstrated no residual tumor. Clinical T stage (cT1-cT3) generally corresponded to pathological T findings. No clear relationship was found between MRI–pathology size discrepancies and either LVSI status (33.3%) or higher tumor grade (40% of patients).
Conclusion: This pilot study demonstrates good correlation between MRI and histopathological measurements. While measurement accuracy showed no significant associations with histological features, these findings support the feasibility of MRI simulation and the potential for MR-linac-guided treatment. Larger prospective studies using standardized protocols are warranted to validate these results and optimize MRI-guided adaptive treatment strategies.