An automatic classification of breast cancer using fuzzy scoring based ResNet CNN model

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Biology

Background:

• The rapid expansion of medical data necessitates efficient automated disease diagnosis systems.
• Existing machine learning (ML) models face challenges in computation time and prediction accuracy.
• Deep learning (DL) offers potential for improving diagnostic efficiency and speed.

Purpose of the Study:

• To develop a hybrid DL model for automated disease diagnosis with improved prediction performance and reduced computation time.
• To enhance classifier accuracy through a novel pre-processing method utilizing statistical co-relational evaluation.
• To optimize the DL model using swarm intelligence for feature selection and reduced processing complexity.

Main Methods:

• A hybrid DL model integrating Fuzzy Scoring Resnet-Convolutional Neural Network (FS-Resnet CNN) was developed.
• Feature extraction employed the wrapping technique and fast discrete wavelet transform (FDWT) from Region of Interest (ROI) images.
• The Adaptive Grey Wolf Optimization Algorithm (AGWOA) was utilized for feature selection and reducing processing time.
• Statistical co-relational evaluation was used for pre-processing to improve classifier accuracy.

Main Results:

• The proposed hybrid DL model demonstrated superior prediction performance and reduced computation time compared to ML models.
• The FS-Resnet CNN framework achieved higher detection rates than existing prediction models.
• The model proved computationally effective, less sensitive to noise, and memory-efficient.
• Performance was validated using metrics including recall, precision, F-measure, and accuracy.

Conclusions:

• The developed hybrid DL framework offers an effective and efficient solution for automated disease diagnosis.
• The integration of AGWOA and FS-Resnet CNN significantly enhances diagnostic accuracy and speed.
• This approach addresses the challenges posed by large medical datasets and the need for rapid, reliable diagnoses.