Purpose: To apply experimentally and further develop a new image interpretation model based on repeated imaging and aimed at improving assessments of technical efficacy and diagnostic accuracy in the detection of small lesions.
Material and Methods: VX2 carcinoma was implanted in the liver of 14 rabbits as two 1.1–1.7 mm3 cores. Magnetic resonance imaging was performed before and 4 days after implantation and then every second day up to the 14th to 20th day. One T2‐weighted sequence (TSE T2) and three T1‐weighted sequences (SE T1, GE T1, and TFL T1) were used. Interpretation was performed stepwise: three readers independently interpreted image sequences chronologically (step 1). Tumors were included at the last examination (step 2). By concurrent interpretation of repeated examinations, the earliest day at which tumors became visible and tumor size were recorded (step 3). Records were corrected (step 4) and autopsy was performed (step 5). Two procedures for use in calculating repeated detection rates of tumors with different magnetic resonance imaging sequences are presented and discussed.
Results: Of 40 macroscopic tumors, 34 were included. They were mainly small (size range SE T1: 1–3 mm, TSE T2: 1.5–5 mm) when they became visible as determined at step 3, which was consistently earlier than observed at step 1. TSE T2, SE T1, and GE T1 did not differ significantly regarding earliest day of detection (step 3), while TFL T1 revealed the tumors later. The initial repeated detection rates were higher with TSE T2 than with the other sequences. Frequency of false positives varied over time, indicating fluctuating criteria for reporting tumors.
Conclusion: A theoretical image interpretation model previously described proved to be applicable for detection of experimental liver tumors. The model was improved by introducing calculations of repeated detection rates for initial image interpretation using an imaging reference standard.
