Deep learning-based prostate cancer diagnosis on MRI with hip prostheses: artifact and sequence effects

The authors aimed to assess how hip prosthesis-induced artifacts affect the diagnostic performance of deep learning (DL) models for prostate cancer detection on MRI and to identify the optimal MRI sequence combinations for analysis. Their findings revealed that DL models showed decreased accuracy in the presence of moderate-to-severe artifacts compared to examinations without prostheses, while traditional radiologist assessments maintained better performance. This study underscores the limitations of DL approaches in this context and emphasizes the importance of expert radiologist interpretation.