How artificial intelligence is changing the way we understand grouped breast microcalcifications?

Abstract Background To explore the transformative role of artificial intelligence (AI) in the detailed analysis and characterization of grouped breast microcalcifications, highlighting its potential to improve the diagnostic accuracy. In the current retrospective analysis, artificial intelligence 2D solution was used to read mammograms showing grouped microcalcifications and interpretation included qualitative and quantitative information to assess suspicion of malignancy. All the included abnormal grouped calcification were pathologically proven. Results The performance of AI in providing abnormality scoring percentage varied depending on the type of morphology that is been evaluated. The coarse heterogeneous morphology had a cutoff value of 69.5%, with a balanced sensitivity 87.5% and specificity of 67.8% ( P = 0.000), the fine pleomorphic morphology had a cutoff value of 87.5%, with an accepted sensitivity 63.83% and high specificity of 86.21% ( P = 0.000). The fine linear descriptor had a higher malignancy rate of 52.94% and the lowest performance in both sensitivity and specificity. Conclusions Despite challenges in achieving balanced performance for grouped microcalcifications, the findings affirm the potential of AI-assisted analysis in enhancing diagnostic precision in detecting malignancy in coarse heterogeneous microcalcifications and identifying benign cases in amorphous morphology type.