| Abstract [eng] |
Background and Objective: Prostate cancer is one of the most common cancers in men, and early diagnosis is critical.Segmentation of cancerous regions in multiparametric MRI is a key step. Deep neural networks, such as nnU-Net, performwell, and incorporating prostate zonal information may further improve accuracy. Methods: This study introduces fourprostate cancer segmentation ensembles that integrate zonal data, compared with a baseline model, which uses zonalinformation as a separate input channel. Ensembles employ specific prostate zone cancer segmentation models trained withthe nnU-Net method. To address variability in manual annotations, a new evaluation metric, the tolerant Dice ScoreCoefficient (DSCτ), is proposed, accounting for ground truth inaccuracies. Results: Ensemble 3 yields the best performance,with a 4.77% higher mean DSC and 6.17% higher mean DSCτ than the baseline. Although the metrics of Ensemble 4 are slightlylower, it reduces false positives by 7.79% and uses fewer models (2 vs. 3), making it more efficient. Furthermore, theapplication of the Conover post hoc test for unreplicated blocked data shows that there is no statistically significantdifference in performance metrics between the results of two ensembles. Thus, Ensemble 4 is the preferred approach forprostate cancer segmentation. Additionally, all ensembles achieve 5.03% to 7.13% higher mean DSCτ values compared to thestandard DSC, confirming the effectiveness of the new metric in handling segmentation uncertainties. Conclusion: Theexperiment results indicate that the proposed Ensemble 4 is the most suitable solution for the prostate cancer segmentationtask. Moreover, the results also indicate that the proposed metric, DSCτ, accounts for ground truth segmentation errors. |
