Related Experiment Video
Updated: Aug 14, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.0K
Fourier transformation based texture analysis for differentiating between hyperplastic polyps and sessile serrated
Geula Klorin1,2,3, Noa Hayat3, Revital Linder2
1Department of Internal Medicine B, Rambam Health Care Campus, Haifa, Israel.
Microscopy Research and Technique
|January 10, 2023
Summary
Automated texture analysis of colorectal polyps using Fourier transforms can accurately distinguish between hyperplastic polyps (HP) and sessile serrated adenomas (SSA), aiding in early cancer detection.
Area of Science:
- Gastroenterology
- Computational Pathology
- Medical Imaging Analysis
Background:
- Colorectal cancer (CRC) is a major global health concern, with the serrated pathway being a significant contributor to its development.
- Distinguishing between benign hyperplastic polyps (HP) and premalignant sessile serrated adenomas (SSA) is challenging due to their morphological similarities, impacting accurate diagnosis and patient management.
- Accurate differentiation is crucial as SSAs possess a notable potential for malignant transformation, unlike HPs.
Purpose of the Study:
- To develop and validate an automated method for morphologically differentiating between hyperplastic polyps (HP) and sessile serrated adenomas (SSA).
- To investigate the efficacy of computerized texture analysis of Fourier-transformed histological images for classifying colonic neoplasia.
- To compare the diagnostic performance of statistical regression and neural network models in classifying HP and SSA.
Main Methods:
- Utilized computerized texture analysis on 30 HP and 58 SSA histological images.
- Applied fast Fourier transformation to images, followed by gray level co-occurrence matrix transformation.
- Extracted four textural variables (entropy, correlation, contrast, homogeneity) for analysis using statistical and neural network (NNET) classification models.
Main Results:
- Statistical regression achieved 95% sensitivity for SSA detection and 80% specificity for HP detection.
- Neural network (NNET) analysis demonstrated superior performance with 100% classification accuracy.
- Fourier-based texture image analysis proved effective in differentiating between HP and SSA.
Conclusions:
- Automated Fourier-based texture analysis offers a reliable method for distinguishing between hyperplastic polyps and sessile serrated adenomas.
- The developed neural network model shows high accuracy in classifying colonic polyps, with significant clinical implications for identifying precancerous lesions.
- This approach has the potential to improve the diagnostic accuracy of colonic neoplasia, facilitating timely intervention for high-risk polyps.
