MobileNet-SVM: A Lightweight Deep Transfer Learning Model to Diagnose BCH Scans for IoMT-Based Imaging Sensors

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Delayed illness identification and treatment lead to preventable deaths globally.
• Early detection of serious conditions like cancer significantly improves patient survival and quality of life.
• The Internet of Medical Things (IoMT) offers potential for efficient, accessible healthcare services.

Purpose of the Study:

• To develop a lightweight deep transfer learning (DTL) model for accurate classification of breast cancer histology (BCH) scans.
• To create a model suitable for resource-constrained IoMT imaging sensors.
• To address the limitations of large, parameter-heavy deep learning models in medical imaging.

Main Methods:

• A novel lightweight DTL model, MobileNet-SVM, was developed by hybridizing MobileNet and Support Vector Machine (SVM).
• The model was trained and tested on the BreakHis v1 400x dataset of BCH images.
• The auto-classification capability of the model was evaluated for diagnostic efficiency.

Main Results:

• The MobileNet-SVM model achieved 100% accuracy on the training dataset.
• The model demonstrated a test accuracy of 91% and an F1-score of 91.35%.
• The proposed model requires minimal computational resources, indicating suitability for IoMT applications.

Conclusions:

• The MobileNet-SVM model offers a highly accurate and efficient solution for BCH scan auto-classification.
• This lightweight model is well-suited for deployment on IoMT imaging equipment, enhancing early disease detection.
• The study highlights the potential of optimized deep learning for improving medical diagnostic capabilities in connected healthcare environments.