Showing 1–3 of 3 results for author: Quihui-Rubio, P C

FAU-Net: An Attention U-Net Extension with Feature Pyramid Attention for Prostate Cancer Segmentation

Authors: Pablo Cesar Quihui-Rubio, Daniel Flores-Araiza, Miguel Gonzalez-Mendoza, Christian Mata, Gilberto Ochoa-Ruiz

Abstract: This contribution presents a deep learning method for the segmentation of prostate zones in MRI images based on U-Net using additive and feature pyramid attention modules, which can improve the workflow of prostate cancer detection and diagnosis. The proposed model is compared to seven different U-Net-based architectures. The automatic segmentation performance of each model of the central zone (CZ… ▽ More This contribution presents a deep learning method for the segmentation of prostate zones in MRI images based on U-Net using additive and feature pyramid attention modules, which can improve the workflow of prostate cancer detection and diagnosis. The proposed model is compared to seven different U-Net-based architectures. The automatic segmentation performance of each model of the central zone (CZ), peripheral zone (PZ), transition zone (TZ) and Tumor were evaluated using Dice Score (DSC), and the Intersection over Union (IoU) metrics. The proposed alternative achieved a mean DSC of 84.15% and IoU of 76.9% in the test set, outperforming most of the studied models in this work except from R2U-Net and attention R2U-Net architectures. △ Less

Submitted 3 September, 2023; originally announced September 2023.

Comments: This paper has been accepted at the 22nd Mexican International Conference on Artificial Intelligence (MICAI 2023)

Assessing the performance of deep learning-based models for prostate cancer segmentation using uncertainty scores

Authors: Pablo Cesar Quihui-Rubio, Daniel Flores-Araiza, Gilberto Ochoa-Ruiz, Miguel Gonzalez-Mendoza, Christian Mata

Abstract: This study focuses on comparing deep learning methods for the segmentation and quantification of uncertainty in prostate segmentation from MRI images. The aim is to improve the workflow of prostate cancer detection and diagnosis. Seven different U-Net-based architectures, augmented with Monte-Carlo dropout, are evaluated for automatic segmentation of the central zone, peripheral zone, transition z… ▽ More This study focuses on comparing deep learning methods for the segmentation and quantification of uncertainty in prostate segmentation from MRI images. The aim is to improve the workflow of prostate cancer detection and diagnosis. Seven different U-Net-based architectures, augmented with Monte-Carlo dropout, are evaluated for automatic segmentation of the central zone, peripheral zone, transition zone, and tumor, with uncertainty estimation. The top-performing model in this study is the Attention R2U-Net, achieving a mean Intersection over Union (IoU) of 76.3% and Dice Similarity Coefficient (DSC) of 85% for segmenting all zones. Additionally, Attention R2U-Net exhibits the lowest uncertainty values, particularly in the boundaries of the transition zone and tumor, when compared to the other models. △ Less

Submitted 8 August, 2023; originally announced August 2023.

Comments: Article accepted at Cancer Prevention through early detecTion (CaPtTion) workshop at MICCAI 2023

Comparison of automatic prostate zones segmentation models in MRI images using U-net-like architectures

Authors: Pablo Cesar Quihui-Rubio, Gilberto Ochoa-Ruiz, Miguel Gonzalez-Mendoza, Gerardo Rodriguez-Hernandez, Christian Mata

Abstract: Prostate cancer is the second-most frequently diagnosed cancer and the sixth leading cause of cancer death in males worldwide. The main problem that specialists face during the diagnosis of prostate cancer is the localization of Regions of Interest (ROI) containing a tumor tissue. Currently, the segmentation of this ROI in most cases is carried out manually by expert doctors, but the procedure is… ▽ More Prostate cancer is the second-most frequently diagnosed cancer and the sixth leading cause of cancer death in males worldwide. The main problem that specialists face during the diagnosis of prostate cancer is the localization of Regions of Interest (ROI) containing a tumor tissue. Currently, the segmentation of this ROI in most cases is carried out manually by expert doctors, but the procedure is plagued with low detection rates (of about 27-44%) or overdiagnosis in some patients. Therefore, several research works have tackled the challenge of automatically segmenting and extracting features of the ROI from magnetic resonance images, as this process can greatly facilitate many diagnostic and therapeutic applications. However, the lack of clear prostate boundaries, the heterogeneity inherent to the prostate tissue, and the variety of prostate shapes makes this process very difficult to automate.In this work, six deep learning models were trained and analyzed with a dataset of MRI images obtained from the Centre Hospitalaire de Dijon and Universitat Politecnica de Catalunya. We carried out a comparison of multiple deep learning models (i.e. U-Net, Attention U-Net, Dense-UNet, Attention Dense-UNet, R2U-Net, and Attention R2U-Net) using categorical cross-entropy loss function. The analysis was performed using three metrics commonly used for image segmentation: Dice score, Jaccard index, and mean squared error. The model that give us the best result segmenting all the zones was R2U-Net, which achieved 0.869, 0.782, and 0.00013 for Dice, Jaccard and mean squared error, respectively. △ Less

Submitted 19 July, 2022; originally announced July 2022.