MCE: Medical Cognition Embedded in 3D MRI feature extraction for advancing glioma staging
Han Xue1,2,3, Huimin Lu1,2,3, Yilong Wang3,4,5
1School of Computer Science and Engineering, Changchun University of Technology, Changchun, Jilin, China.
Plos One
|May 31, 2024
Summary
This study introduces a novel Medical Cognition Embedded (MCE) model for 3D brain glioma staging using MRI data. The MCE model enhances feature extraction by integrating medical knowledge, improving diagnostic accuracy over purely data-driven methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Data-driven deep learning for brain glioma MRI classification faces challenges due to limited data and difficulties in feature extraction.
- Current 2D analysis methods require slice selection, which is impractical for clinical use.
Purpose of the Study:
- To propose a novel Medical Cognition Embedded (MCE) model for 3D brain glioma staging.
- To enhance feature extraction quality by embedding medical knowledge into data-driven approaches.
Main Methods:
- Developed a 3D MCE model incorporating deep feature extraction mimicking medical professionals' interpretation of MRI sequences.
- Performed extensive Radiomics feature extraction (texture, morphology, grayscale).
- Created diagnostic features (Diag-Features) by referencing radiological diagnosis and imaging characteristics.
Main Results:
- The MCE model demonstrated superior performance on the BraTS2018 and BraTS2020 datasets.
- Achieved higher accuracy (96.14%, 97.57%), recall (93.4%, 92.80%), and precision (97.06%, 95.96%) compared to purely data-driven models.
Conclusions:
- The proposed MCE model effectively enhances brain glioma staging using 3D MRI data.
- Integrating medical knowledge into AI models improves feature extraction and diagnostic performance in neuro-oncology.
