Three-dimensional breast imaging using Artificial-Intelligence-Based Automatic Measurement System.

Background: Three-dimensional (3D) image technology in breast measurement requires exploration. We aimed to evaluate a new automatic breast measurement system based on artificial intelligence (AI).

Methods: This prospective controlled study included all-women patients who underwent breast reconstruction from January to May 2022. Patients underwent 3D scanning before breast reconstruction. Two doctors performed the measurements twice through AI and manual measurements on the 3D images, respectively. The measurement results of bilateral breast width, convexity, height, volume, and measurement time were recorded. Consistency analyses were performed.

Results: Fifty-eight patients (116 breasts) were recruited. For the left breasts, AI and manual measurements showed excellent consistency (intra-class correlation coefficients (ICC) = 0.81) in width measurements, moderate consistency (ICC = 0.59) in height measurements, excellent consistency (ICC = 0.87) in convexity measurements, and good consistency (ICC = 0.74) in volume measurements. For the right breasts, the width consistency was excellent (ICC = 0.93), height consistency was good (ICC = 0.65), convexity consistency was excellent (ICC = 0.94), and volume consistency was excellent (ICC = 0.85). The Bland-Altman curves also showed that the measurement results were comparable and few outliers were detected. AI average measurement time (compared to manual measurements) was significantly shorter (40.65 ± 1.51 s vs. 610.47 ± 18.74 s; p < 0.001).

Conclusion: The AI-based 3D breast measurement system showed high accuracy, better reproducibility, and significantly shortened the measurement time, which could help guide surgical management.