Correlation Between Imaging and Pathologic Tumor Delineation for Non-colorectal Hepatic Metastases

Abstract

The liver is a common site of distant metastases for many different primary malignancies. The appearance of hepatic metastases is variable on computed tomography (CT), and may present as hypoattenuating or can have peripheral enhancement with contrast. Magnetic resonance imaging (MRI) has many advantages to CT imaging, but the prevalent imaging modality for linear accelerators is CT-based. We aim to investigate the concordance between imaging and pathologic size in patients with resected hepatic metastases to determine the appropriateness of CT and MRI for stereotactic body radiation therapy (SBRT) gross tumor volume (GTV) delineation. An institutional database of patients with resected hepatic metastases was queried for patients with metastatic disease from non-colorectal primary malignancies including breast, neuroendocrine, melanoma, ovarian, biliary, and gastrointestinal stromal tumors. Patients with either preoperative MRI or CT at our institution were selected for the study. Largest tumor dimension was obtained from each preoperative MRI or CT, as well as from the postoperative pathologic tumor specimen. A general linear regression model was used to model the outcome largest dimension from imaging modality minus largest dimension from pathology. The study identified 73 patients with preoperative imaging and gross pathologic specimen measurements; 32 patients had preoperative CT and 41 patients had preoperative MRI. Median time in days between preoperative imaging and hepatectomy was 20 (range 2-78) for CT, and 13 (range 1-80) for MRI. The median largest dimension on pathologic specimen was 5.5 cm (range 1.3 -25) for patients with preoperative CT, and 3.6 cm (range 0.3 -25) for patients with preoperative MRI. When compared to pathologic tumor dimensions, CT underestimated tumor size by a mean of 9.06mm (p=0.013), and MRI underestimated tumor size by a mean of 3.78mm (p=0.032). Upon comparing MRI and CT, MRI demonstrated a trend for better tumor size delineation with concurrence correlation coefficient (CCC) 0.98 (95% confidence interval (CI); 0.96, 0.99) compared to CCC 0.92 for CT (95% CI; 0.86, 0.96). Larger tumor size was also found to be underestimated more (p<0.001) amongst all patients. Both CT and MRI underestimate the pathologic size of resected non-colorectal liver metastases with the greatest underestimation in larger tumors. There was a trend towards better delineation with MRI. Target delineation for liver SBRT is challenging, and the determination of optimal imaging for treatment simulation and planning remains debatable. Our data suggest that the imaging modality used for treatment planning should be considered when determining the final GTV margin.