Prostate lesion segmentation from MR images using deep learning methods
Keywords:Deep Learning, Prostate MRI, DeepLabV3 , Lesion Detection
Prostate cancer is a prevalent form of cancer in men, emphasizing the need for accurate and efficient methods for prostate lesion segmentation from magnetic resonance (MR) images. Manual segmentation of prostate lesions is time-consuming and subjective, highlighting the significance of automated approaches using deep learning methods. This study presents a comprehensive investigation into applying deep learning techniques for prostate lesion segmentation from MR images. The study explores state-of-the-art deep learning models, including U-Net, PAN, DeepLabV3, DeepLabV3+ for segmentation. A large PICAI dataset of prostate MR images, comprising multi-parametric MRI scans and expert annotations, is utilized for evaluating the developed methods. Performance metrics such as accuracy, precision, recall, specificity, accuracy, IOU, AP Score, PR curve, and AUC curve were employed to compare the proposed deep learning methods. In summary, this research contributes to the field of prostate lesion segmentation by investigating the effectiveness of deep learning methods applied to MR images. The DeepLabV3+ model achieves an IOU of 0.79 and an AP of 0.54 using Jaccard Loss.
Siegel, R. L., Miller, K. D., Wagle, N. S., & Jemal, A. (2023). Cancer statistics, 2023. CA: a cancer journal for clinicians, 73(1), 17-48.
Karagoz, A., Seker, M. E., Yergin, M., Kan, T. A., Kartal, M. S., Karaarslan, E., ... & Oksuz, I. (2023). Prostate Lesion Estimation using Prostate Masks from Biparametric MRI. arXiv preprint arXiv:2301.09673.
Saha, A., Bosma, J., Twilt, J., van Ginneken, B., Yakar, D., Elschot, M., ... & de Rooij, M. (2023, April). Artificial Intelligence and Radiologists at Prostate Cancer Detection in MRI—The PI-CAI Challenge. In Medical Imaging with Deep Learning, short paper track.
Buslaev, A., Iglovikov, V. I., Khvedchenya, E., Parinov, A., Druzhinin, M., & Kalinin, A. A. (2020). Albumentations: fast and flexible image augmentations. Information, 11(2), 125.