Abbreviated Liver MRI

Introduction

Colorectal cancer is one of the most common malignancies worldwide, and the liver is both the first and most common site of solid organ metastasis. Therefore, accurate detection of liver metastases is critical for treatment planning and surgical decision-making.

While gadoxetate-enhanced MRI (Primovist/Eovist) offers excellent sensitivity, its lengthy protocol limits throughput and patient tolerance. This study explores whether an abbreviated MRI protocol can maintain diagnostic accuracy while significantly reducing scan time.

Study Overview

• Study type: Prospective, randomized controlled non-inferiority trial
• Design: Multi-reader, multi-case; balanced group assignment; allocation concealed; unblinded
• Population: 200 adult patients with colorectal cancer and known or suspected liver metastases
• Comparison: Abbreviated protocol (T2, DWI/ADC, and hepatobiliary phase) vs. full Primovist protocol
• Readers: Seven radiologists at London Health Sciences Centre (LHSC) and Vanderbilt University Medical Center (VUMC)
• Reference standard: Histopathology when available, otherwise imaging follow-up
• Follow-up: One year, used to establish the reference standard and track clinical outcomes and patient healthcare utilization
• Primary outcome: Per-segment diagnostic accuracy for liver metastasis detection, measured by AUROC
• Secondary outcomes: Downstream healthcare utilization; overall survival; interobserver variability; reading time

This trial is registered at ClinicalTrials.gov (NCT05314400).

Team

• Principal Investigator: Dr. Harry Marshall
• Research Coordinator: Karen Dunn
• Study Conception: Dr. Zahra Kassam, Dr. Narinder Paul
• Readers: Dr. Darryl Ramsewak, Dr. Vibhuti Kalia, Dr. Virginia Planz, Dr. Krupa Patel-Lippmann, Dr. Ahmed Mohamed, Dr. Katherine Frederick-Dyer, Dr. Jennifer Huang
• Surgical Collaborators Dr. Anton Skaro, Dr. Ephraim Tang, Dr. Julie Ann Van Koughnett, Dr. Douglas Quan
• Imaging Data Collection: Dr. Shelby Champion
• Pathology Data Collection: Natalie Grindrod
• Lead MRI Technologist: Douglas Cesarin
• AI Development: Dr. Aaron Ward, Christina Lim, Carol Johnson, Robert Policelli

Preliminary Findings

A preliminary analysis was performed using three of the seven planned abdominal radiologist readers on 58 patients (35 with metastases, 23 without), including a total of 367 lesions. Of these, 220 lesions (113 metastatic, 107 benign) identified by all three readers were included in the lesion-level accuracy analysis.

Per-lesion diagnostic performance was high for both protocols (Figure 1). Mean AUROC was 0.93 (range: 0.89–0.96) for the abbreviated protocol and 0.96 (range: 0.92–0.99) for the full protocol. The observed mean AUROC was 0.03 lower for the abbreviated protocol than for the full protocol. These descriptive results are preliminary; a conclusion about non-inferiority will require the planned statistical analysis of the complete reader dataset. Patient-level AUROC results showed a similar trend, with mean AUCs of 0.88 (abbreviated) and 0.91 (full).

Full statistical testing will be performed once all reader data are available.

Figure 1: Per-lesion ROC curves comparing abbreviated (green) and full (yellow) protocols for liver metastasis detection.

Implications

These preliminary observations describe diagnostic performance in the readers evaluated so far. Analysis of the complete dataset will assess the abbreviated protocol’s performance relative to the full protocol, together with reading time, interobserver variability, and downstream healthcare utilization. The study will help clarify whether a shorter liver MRI protocol could be useful in practice.

Next steps

Next steps are to complete data collection from all seven readers, finalize follow-up data on overall survival and healthcare utilization, and perform the planned analyses of diagnostic performance, interobserver variability, and reading time.