Related Experiment Video
Updated: Oct 11, 2025
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
378
Deep Learning-Based Diagnosis Method of Emergency Colorectal Pathology
1Pathology Department, Wuhan Hanyang Hospital, Wuhan, Hubei 430050, China.
Journal of Healthcare Engineering
|November 29, 2021
Summary
Emergency colorectal cancer diagnosis needs computer assistance. This study presents a gland segmentation and deep learning approach for accurate pathology image analysis, improving diagnostic efficiency and reducing errors in emergency settings.
Area of Science:
- Medical Imaging
- Computational Pathology
- Oncology
Background:
- Emergency colorectal cancer presents a growing challenge with high morbidity and mortality rates.
- Current diagnostic methods heavily rely on pathologist experience, leading to potential misdiagnosis and heavy workloads.
- There is a critical need for computer-aided diagnosis (CADx) systems to enhance accuracy and efficiency in emergency colorectal pathology.
Purpose of the Study:
- To develop and evaluate a computer-aided diagnosis system for emergency colorectal pathology images.
- To improve the accuracy and efficiency of distinguishing between benign and malignant colorectal pathologies.
- To explore the utility of gland segmentation and deep learning models in this diagnostic context.
Main Methods:
- A multi-featured auxiliary diagnosis model using Support Vector Machines (SVM) based on contour, color, and texture features was developed.
- Glandular segmentation was applied to pathological images to create a refined dataset (D2) for comparison with original images (D1).
- Deep convolutional neural networks (CNNs), specifically CIFAR and VGG, were implemented and trained for pathology diagnosis.
Main Results:
- The SVM model achieved higher diagnostic accuracy on the segmented dataset (D2).
- The highest accuracy of 83.75% was obtained using an SVM model combining contour and texture features on D2, indicating limitations of traditional methods.
- Deep convolutional neural networks (CIFAR and VGG) were successfully trained and applied to both original (D1) and segmented (D2) datasets for diagnosis.
Conclusions:
- Computer-aided diagnosis, particularly using deep learning and image segmentation, shows significant promise for improving emergency colorectal cancer diagnosis.
- Glandular segmentation enhances the performance of diagnostic models.
- Further development of advanced computational methods is crucial for addressing the challenges in emergency colorectal pathology diagnosis.
