Related Experiment Video
Updated: Aug 6, 2026
10:37
A Multimodal Imaging Framework to Advance Phenotyping of Living Label-free Breast Cancer Cells
Published on: August 22, 2025
Multimodal breast cancer diagnosis using feature fusion and deep learning
Varun Malik1, Tahani Alsubait2, Mudassir Khan3
1Chitkara University Institute of Engineering and Technology, Chitkara University, Rajpura, India.
Frontiers in Medicine
|August 5, 2026
Summary
This study introduces a novel multimodal breast cancer diagnosis model using advanced AI techniques. The model achieves high accuracy, demonstrating its potential for improved early detection and patient outcomes in diverse clinical settings.
Area of Science:
- Oncology
- Artificial Intelligence
- Medical Imaging
Background:
- Breast cancer diagnosis relies heavily on accurate and timely detection, which is challenged by data limitations and disease complexity.
- Traditional deep learning models struggle with single-modality data, hindering comprehensive analysis of heterogeneous breast cancer characteristics.
- Incomplete or unavailable data further complicates accurate breast cancer diagnosis using conventional methods.
Purpose of the Study:
- To develop a robust multimodal breast cancer diagnosis model capable of handling data complexity and incompleteness.
- To enhance the accuracy and efficiency of breast cancer diagnosis through advanced feature extraction and fusion techniques.
- To create a generalizable and resilient diagnostic tool for diverse clinical imaging data.
Main Methods:
- Utilized attention-based transformers for efficient, modality-specific feature extraction.
- Employed the modified mantissa search (MMS) algorithm to eliminate irrelevant features.
- Integrated the American zebra optimization (AZO) algorithm for dynamic and efficient feature combination.
- Implemented a lightweight convolutional neural network (LCNN) for final classification to maintain diagnostic accuracy.
Main Results:
- Achieved high diagnostic accuracy across multiple datasets: 98.958% on MIAS, 97.37% on BreakHis, and 99.438% on combined multimodal data.
- Demonstrated exceptional generalizability and resilience to missing data modalities.
- Validated the model's effectiveness in clinical diagnostic scenarios involving varied imaging data.
Conclusions:
- The proposed multimodal diagnostic model significantly improves breast cancer detection accuracy and reliability.
- The model's ability to handle missing data makes it a valuable tool for real-world clinical applications.
- This approach offers a promising advancement in leveraging multimodal data for enhanced breast cancer diagnostics.