Prostate MRI deep learning algorithm robust to reconstruction-accelerated inputs but with fewer false positives

A prostate cancer DLA tested on DLR-accelerated bpMRI showed similar per-patient performance. However, with C-T2WI+DLR-DWI, detected lesions fell (99 vs 131), FP/case dropped (0.478 vs 0.722, P=0.003), and IoU improved (0.194 vs 0.153, P=0.014).

• Retrospective study in 141 men (115 with definitive diagnosis) compared the DLA across four biparametric MRI configurations combining conventional and DLR-accelerated T2WI and DWI.
• The DLA maintained largely comparable diagnostic performance across configurations, but using DLR-DWI led to fewer detected lesions and significantly lower false positives per case.
• The vendor-specific nature of both the deep learning algorithm and the reconstruction limits generalizability and warrants further validation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.