Cervical cancer diagnosis based on modified uniform local ternary patterns and feed forward multilayer network
Shervan Fekri-Ershad1, S Ramakrishnan2
1Faculty of Computer Engineering, Najafabad Branch, Islamic Azad University, Najafabad, Iran; Big Data Research Center, Najafabad Branch, Islamic Azad University, Najafabad, Iran.
Computers in Biology and Medicine
|March 17, 2022
Summary
This study introduces a novel two-stage method for classifying cervical cancer in pap smear images using modified uniform local ternary patterns (MULTP) and a genetically optimized neural network. The new approach achieves higher detection accuracy and is efficient for online diagnosis.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Cervical cancer is a leading cause of cancer death in women, necessitating accurate and early diagnosis.
- Pap smear testing is a common diagnostic tool, relying on visual analysis of cervical cell morphology.
- Existing image processing methods for pap smear analysis often prioritize accuracy but may lack efficiency or robustness.
Purpose of the Study:
- To present a two-stage image processing method for automated cervical cancer diagnosis from pap smear images.
- To introduce a novel texture descriptor, modified uniform local ternary patterns (MULTP), for feature extraction.
- To develop a genetically optimized multi-layer feed-forward neural network for improved classification accuracy.
Main Methods:
- Image segmentation using optimal thresholding to isolate relevant cellular regions.
- Application of the proposed MULTP descriptor to extract joint texture information from cytoplasm and nucleolus.
- Classification using a deep neural network optimized via a genetic algorithm for network architecture (hidden layers and nodes).
Main Results:
- The proposed method demonstrated higher detection accuracy compared to existing methods on the Herlev database.
- The method exhibits insensitivity to image rotation, a significant advantage for practical applications.
- Low run time indicates suitability for online diagnostic systems.
Conclusions:
- The presented two-stage method offers a robust and accurate approach for cervical cancer detection in pap smear images.
- The MULTP texture descriptor is a versatile tool applicable to various computer vision texture analysis tasks.
- The genetic algorithm optimization technique can enhance the performance of deep neural networks in medical image analysis.
