- Collaborated with the Pediatrics Department of NTUH Hsinchu Branch.
- Utilized near-infrared spectroscopy to measure prefrontal cortex oxygenation changes in pediatric patients with common mental disorders during testing.
- Constructed a CNN model for multi-class classification, distinguishing between healthy and diseased children. Achieved an accuracy rate of approximately 80%.
- Applied explainable AI to visualize the results.
- In this study, we apply neurofeedback training based on functional near-infrared spectroscopy as the behavior therapy for TS patients, and last for 8 weeks.
- Yale global tic severity scale (YGTSS) is applyed in the first and last week to evaluate the improvment of patients, the mild group and severe group is bounded by score of 25.
- The SVC helps us the figure out the differences between mild group and severe group.
Public Rank | Public Score | Private Rank | Private Score |
---|---|---|---|
5 / 307 | 0.918 | 6 / 307 | 0.916 |
- Encoder-decoder structure
- Encoder: ECA-NFNet-l1 & PAN
- Decoder: EfficientNetV2-s & DeepLabV3+
- Ensemble voting
Public Rank | Public Score | Private Rank | Private Score |
---|---|---|---|
6 / 371 | 0.659922 | 28 / 371 | 0.558209 |
- Two input model: MLP & CNN
- Use MFCC to deal with acoutics data
- Python
- Matlab
- C++
- LabVIEW
- Windows
- Linux
- sci-kit learn
- pytorch
- NumPy
- Pandas
- Matplotlib
- Seaborn