Related Experiment Video
Updated: Oct 26, 2025
11:27
Quantitative Multispectral Analysis Following Fluorescent Tissue Transplant for Visualization of Cell Origins, Types, and Interactions
Published on: September 22, 2013
9.5K
A neural pathomics framework for classifying colorectal cancer histopathology images based on wavelet multi-scale
Eleftherios Trivizakis1,2, Georgios S Ioannidis3, Ioannis Souglakos4,5
1Medical School, University of Crete, 71003, Heraklion, Greece. trivizakis@ics.forth.gr.
Scientific Reports
|July 31, 2021
Summary
An artificial neural network accurately classifies colorectal cancer (CRC) tissue pathology. This automated system achieved 95.3% accuracy, improving diagnostic precision for CRC detection.
Area of Science:
- Oncology
- Computational Pathology
- Artificial Intelligence
Background:
- Colorectal cancer (CRC) is a leading cause of cancer-related deaths globally.
- Accurate histopathological classification of CRC is crucial for diagnosis and treatment planning.
- Automated systems for CRC tissue classification can enhance diagnostic accuracy and reduce pathologist workload.
Purpose of the Study:
- To develop and evaluate an artificial neural network for automated classification of CRC tissue regions.
- To investigate the impact of multi-level pathomics features and scale on CRC tissue differentiation.
- To compare the performance of the proposed model against existing studies.
Main Methods:
- Training an artificial neural network on 5000 CRC histopathology image tiles.
- Extracting 532 multi-level pathomics features using visual descriptors (e.g., local binary patterns, wavelet transforms, Gabor filters).
- Performing exhaustive evaluation with various wavelet families and parameters, including tenfold cross-validation.
Main Results:
- The developed artificial neural network achieved a classification accuracy of 95.3%.
- This performance significantly surpasses the 87.4% accuracy reported in recent studies.
- The study demonstrated that the first and second levels of wavelet approximations can be utilized without compromising classification performance.
Conclusions:
- The proposed automated system demonstrates high accuracy in classifying CRC tissue pathology.
- The findings suggest that specific wavelet approximation levels can be effectively used for feature extraction, optimizing the process.
- This AI-driven approach holds significant potential for improving diagnostic precision and efficiency in colorectal cancer management.
