Related Experiment Video
Updated: Oct 4, 2025
06:08
Quantitative Visualization and Detection of Skin Cancer Using Dynamic Thermal Imaging
Published on: May 5, 2011
17.0K
BreaCNet: A high-accuracy breast thermogram classifier based on mobile convolutional neural network.
Roslidar Roslidar1,2,3, Mohd Syaryadhi2, Khairun Saddami2
1Doctoral Program, School of Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia.
Mathematical Biosciences and Engineering : MBE
|February 9, 2022
Summary
A novel mobile deep learning model, BreaCNet, achieves 100% accuracy for early breast cancer detection using thermal imaging. This mobile neural network (CNN) enhances diagnostic sensitivity and specificity, prioritizing user data privacy through on-device processing.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Early breast cancer detection is crucial for improved patient outcomes.
- Mobile-based diagnostic tools require accurate and efficient deep learning models.
- Thermography offers a non-invasive imaging modality for breast cancer screening.
Purpose of the Study:
- To develop a mobile neural network (BreaCNet) for early breast cancer detection using breast thermograms.
- To create an effective segmentation algorithm for identifying the region of interest (ROI) in thermograms.
- To implement a mobile convolutional neural network (CNN) classifier with high accuracy and efficiency.
Main Methods:
- Developed BreaCNet, integrating a segmentation algorithm (edge detection, polynomial fitting) with a modified ShuffleNet classifier.
- Modified ShuffleNet by adding a convolutional layer with 1028 filters, resulting in 6.1 million parameters and a 22 MB model size.
- Evaluated the model's performance using simulation results, focusing on accuracy, sensitivity, and specificity.
Main Results:
- The modified ShuffleNet classifier alone achieved 72% accuracy.
- Integrating the segmentation algorithm with the classifier boosted accuracy to 100%.
- BreaCNet significantly improved diagnostic accuracy, increasing sensitivity from 43% to 100% and specificity to 100%.
Conclusions:
- BreaCNet demonstrates high diagnostic accuracy for early breast cancer detection using mobile thermography.
- Utilizing the ROI through segmentation enhances classifier performance.
- On-device inference is recommended for BreaCNet to ensure data privacy and reliable performance.
