A training program to reduce reader search errors for liver metastasis detection in CT

Area of Science:

• Medical Imaging
• Radiology
• Hepatology

Background:

• Detection of low contrast liver metastases is inconsistent among radiologists, leading to variability in patient care.
• Targeted training interventions may enhance radiologist performance and reduce inter-reader discrepancies in identifying hepatic lesions.

Purpose of the Study:

• To evaluate the impact of a structured training program on radiologist performance in detecting liver metastases using eye-tracking technology.
• To differentiate between search and classification errors to identify specific areas for training improvement.

Main Methods:

• Thirty-one radiologists (trainees and staff) underwent four reading sessions: pre-test, search training, classification training, and post-test.
• Eye-tracking monitored interpretation, with training focusing on search strategies and lesion classification using CT exams and image patches.
• Missed metastases were categorized as search errors (short gaze time) or classification errors (long gaze time).

Main Results:

• Overall sensitivity for detecting liver metastases increased by 2.8% (p = 0.01) post-training, while AUC remained stable.
• Search errors significantly decreased from 10.8% to 8.1% (p < 0.01), with greater improvement seen in detecting easier metastases.
• Classification errors showed no significant change, remaining at 5.7%.

Conclusions:

• The implemented training program effectively improved radiologist sensitivity by reducing search-related errors in liver metastasis detection.
• The training did not significantly impact classification accuracy, suggesting a need for different strategies to address this type of error.
• Future training should focus on enhancing both search efficiency and classification precision for comprehensive improvement in liver lesion detection.