Implementation and Application Evaluation of a Medical Image Structure Reporting System

Objective To implement imaging diagnostic guidelines and enhance report standardization, we developed a multimodal structured reporting system.

Methods The system integrated "standard coding + structured items + key images" with disease-specific templates, generating text-only and image-text-based reports. A comparative analysis against traditional reports evaluated efficiency, acceptability, completeness of disease feature expression, and data classification accuracy.

Results Image-text-based reports showed significantly higher quality than traditional reports (P<0.01), while text-only reports had no significant difference (P>0.01). Image-text-based reports also surpassed both traditional and text-only reports in information completeness and guideline adherence. Senior radiologists (4.04±0.55) and clinicians (4.19±0.58) reported higher acceptance than junior radiologists (3.04±1.55). For data classification, NLP-based retrieval achieved superior accuracy (F1-Score:0.85~1.00) over keyword-based methods.

Conclusion Image-text-integrated structured reporting reduces heterogeneity in conventional reports and aids competency development among junior radiologists in primary care.