An efficient segmentation of multi-class MRI prostate region by boundary-matching with deep autoencoder Unet model

Prostate cancer is one of the most common types of cancers to affect men worldwide. The cancer area must be precisely segmented from the magnetic resonance imaging (MRI) scan to enhance the efficiency of treatments and the patient’s chances of overcoming prostate cancer (PC). However, identifying tumor locations within the prostate gland poses significant challenges due to variations in prostate shapes among patients, indistinct boundary delineation, and the utilization of multiple MRI modes. To solve these issues, a novel 3D U-Net Deep Autoencoder approach is used to predict and detect prostate tumor locations within MRI datasets. The models encoder block is exploited for feature extraction from input images and expeditious Region Of Interest selection through AutoEncoder training. Furthermore, the Block-Matching and 3D Filtering and histogram equalization techniques are employed to enhance grayscale intensity values at specific ROI boundaries in PC images. Moreover, the implementation of k-fold cross-validation ensures a robust evaluation of the training model. The proposed methodology has been validated using the Medical Segmentation Decathlon Prostate1 dataset for detecting PC Segmentation. To validate the proposed approach several performance evaluation metrics are used, namely, accuracy, Jaccard Similarity Coefficient, Mean Square Error, Dice Similarity Coefficient, Structural Similarity Index Measure, Hausdorff Distance, Peak Signal to Noise Ratio, Average Boundary Distance, and Intersection Over Union to evaluate the experimental outcomes. The proposed scheme outperforms performance compared with state-of-the-art segmentation models when applied to the MRI prostate cancer dataset.

Authors

Institutions

Publication Details

Journal
Evolving Systems
Published
2026-11-18
DOI
https://doi.org/10.1007/s12530-026-09893-z
Primary Topic
Medical Image Segmentation Techniques
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An efficient segmentation of multi-class MRI prostate region by boundary-matching with deep autoencoder Unet model

Piyush Kumar, Anita Murmu, Sangeeta Kumari
3 links
Evolving Systems
Medical Image Segmentation Techniques
article

An efficient segmentation of multi-class MRI prostate region by boundary-matching with deep autoencoder Unet model

Piyush Kumar, Anita Murmu, Sangeeta Kumari
article en

Abstract

Prostate cancer is one of the most common types of cancers to affect men worldwide. The cancer area must be precisely segmented from the magnetic resonance imaging (MRI) scan to enhance the efficiency of treatments and the patient’s chances of overcoming prostate cancer (PC). However, identifying tumor locations within the prostate gland poses significant challenges due to variations in prostate shapes among patients, indistinct boundary delineation, and the utilization of multiple MRI modes. To solve these issues, a novel 3D U-Net Deep Autoencoder approach is used to predict and detect prostate tumor locations within MRI datasets. The models encoder block is exploited for feature extraction from input images and expeditious Region Of Interest selection through AutoEncoder training. Furthermore, the Block-Matching and 3D Filtering and histogram equalization techniques are employed to enhance grayscale intensity values at specific ROI boundaries in PC images. Moreover, the implementation of k-fold cross-validation ensures a robust evaluation of the training model. The proposed methodology has been validated using the Medical Segmentation Decathlon Prostate1 dataset for detecting PC Segmentation. To validate the proposed approach several performance evaluation metrics are used, namely, accuracy, Jaccard Similarity Coefficient, Mean Square Error, Dice Similarity Coefficient, Structural Similarity Index Measure, Hausdorff Distance, Peak Signal to Noise Ratio, Average Boundary Distance, and Intersection Over Union to evaluate the experimental outcomes. The proposed scheme outperforms performance compared with state-of-the-art segmentation models when applied to the MRI prostate cancer dataset.

Evolving SystemsVol. 17(4)
National Institute of Technology Patna (IN), University of Malta (MT), National Institute of Technology Raipur (IN)
Università ta' Malta
Good health and well-being
Medical Image Segmentation Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon