專題報告12/17
MDD/NC
流程
資料
資料處理
EEG Net
訓練函數
MDD/NC
分類結果
EEG Net
模型分析
資料
| Train/Val | Test | |
| MDD | 24 | 5 |
| NC | 20 | 5 |
資料處理
- SSS
- Resample 128 Hz
- Filter 2-30 Hz
- Pick Mag
- Crop 2 sec Overlap 1 sec
- Hanning Window
Before
After
訓練函數
- Normalize
- Channel Drop
- EEG Net
- Binary Cross Entropy
- Adam
- Learning Rate = 1e-3
- 10 Epochs
EEG Net
- Kernel 0: (1,64)
- Kernel 1:(306,1)
- Pool (4)
- Kernel 2:(1,16)
- Kernel 3:(1,1)
- Pool (4)
結果
- Testing: 71\(\rightarrow\)75%
- Training: 77%
- Validation: 76%
Single Trial
Patients
| True | False | |
| MDD | 22/4 | 4/1 |
| NC | 16/4 | 2/1 |
結果
EEG Net
rTMS預測
流程
資料
資料處理
訓練函數
+ / -
分類結果
模型分析
EEG Net
Regression
資料
| Train/Val | Test | |
| Positive | 10 | 4 |
| Negtive | 11 | 4 |
資料處理
- SSS
- Resample 128 Hz
- Filter 2-30 Hz
- Pick Mag
- Crop 2 sec Overlap 1 sec
- Hanning Window
訓練函數
- Normalize
- Channel Drop
- EEG Net
- Binary Cross Entropy
- Adam
- Learning Rate = 1e-3
- 10 Epochs
EEG Net
- FFT
- Normalize
- Channel Drop
- Linear Regression
- Binary Cross Entropy
- Adam
- Learning Rate = 1e-3
- 10 Epochs
Regression
結果
- Testing: 38%
- Training: 99%
- Validation: 87%
EEG Net
Regression
- Testing: 45%
- Training: 80%
- Validation: 80%
Paper Review
Machine learning in major depression: From classification to treatment outcome prediction
fMRI
Neural Networks based approaches for Major Depressive Disorder and Bipolar Disorder Diagnosis using EEG signals: A review
EEG
問題們
- rTMS
- Visualizations
