This is a my Bachelor Diploma work at Igor Sikorsky Kyiv Polytechnic Institute.
I have trained models on XGBoost for each disease images.
Available diseases images:
| Disease | Image Count |
|---|---|
| Dyscholia | 5 |
| Hepatitis B | 9 |
| Hepatitis C | 11 |
| Wilson's Disease | 6 |
| Autoimmune Hepatitis | 8 |
| Norma | 68 |
As you see, amount of disease images are pretty low, to build some models, but I tried anyway. If you have some Ultrasound Images to share with, please contact me.
When you input some image, pre-trained models predicts output with GLRLM features and show percentage output for each
- GLCM
- GLRLM
GLCM Accuracy:
| Dataset | Predictor | Accuracy |
|---|---|---|
| GLCM_0 | XGBoost | 50-60% |
GLRLM Accuracy on Binary Classification:
| Dataset | Predictor | Accuracy |
|---|---|---|
| GLRLM_0 | XGBoost | 88-96% |
| GLRLM_0 | LogisticReg | 85-95% |
| GLRLM_0 | SVM | 86-93% |
| GLRLM_0 | KNN | 86-93% |
python3 main.py image.pngThese are examples of Gradient Boosting model probability prediction
python3 main.py data/images/png/norm/norma0.pngNorma: 98%
Autoimmune Hepatitis: 0%
Dyscholia: 0%
Hepatitis B: 0%
Hepatitis C: 0%
Wilson's Disease: 0%
python3 main.py data/images/png/dsh/4.pngNorma: 2%
Autoimmune Hepatitis: 3%
Dyscholia: 85%
Hepatitis B: 2%
Hepatitis C: 2%
Wilson's Disease: 2%
python3 main.py data/images/png/auh/6.pngNorma: 1%
Autoimmune Hepatitis: 92%
Dyscholia: 1%
Hepatitis B: 0%
Hepatitis C: 1%
Wilson's Disease: 1%
python3 main.py data/images/png/hpb/1.pngNorma: 15%
Autoimmune Hepatitis: 0%
Dyscholia: 0%
Hepatitis B: 83%
Hepatitis C: 0%
Wilson's Disease: 0%
python3 main.py data/images/png/hpc/1.pngNorma: 0%
Autoimmune Hepatitis: 2%
Dyscholia: 0%
Hepatitis B: 0%
Hepatitis C: 94%
Wilson's Disease: 3%
python3 main.py data/images/png/wls/8.pngNorma: 5%
Autoimmune Hepatitis: 1%
Dyscholia: 0%
Hepatitis B: 0%
Hepatitis C: 1%
Wilson's Disease: 90%
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent