| This study proposes an automated segmentation of prostate cancer (PCa) in transrectal ultrasound (TRUS) images using different preprocessing methods to enhance the segmentation accuracy. We propose the use of image intensity nor-malization and despeckle filtering, individually and in combination, as prepro-cessing techniques to improve the performance of a deep learning (DL) segmen-tation model (DeepLabv3+) in ultrasound images of PCa. This algorithm was ap-plied to a dataset of 647 TRUS images. All images were separated into four groups as follows: original (O), intensity normalized (N), despeckled (D), and in-tensity normalized and despeckled (ND). Manual segmentations of the prostate were performed by an experienced radiation oncologist and compared with auto-mated segmentations using six different evaluation metrics. Statistical analysis showed that preprocessing enhances segmentation performance, with a medi-an(±IQR) Dice coefficient (DC) of 94.02(3.93)/94.84(3.92)/94.43(3.05)/94.22(4.19) for the O/N/D/ND images re-spectively. The highest segmentation accuracy was achieved on the N images, followed by the ND images which confirm the benefits of N and ND in enhanc-ing the final segmentation accuracy. No statistically significant differences were found between all different preprocessing schemes for all the evaluation metrics investigated. Additional experimentation with a larger image dataset and other al-ternative evaluation metrics is required to validate the present results. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.