Deep learning based frecture assessment .ppt

Fracture Risk Assessment

1. Fracture Risk
Assessment

2. osteoporosis and bone fracture prediction
Risk factor analysis
Electronic health records (EHRs)
Faster bone loss, demographics, family history,
lifestyle and others
Features selection
Relationship between a disease and RF
FRAX for data processing
FRAX need further interpretation
Prediction and Informative Risk Factor
Selection of Bone Diseases
Li, Hui, et al. "Prediction and informative risk factor selection of bone
diseases." IEEE/ACM Transactions on Computational Biology and
Bioinformatics (TCBB) 12.1 (2015): 79-91

3. Predict unknown sample
based one diseased and un-
diseased patients
DBN learning algorithm
Bone disease prediction
based on integrated features
Informative RF selection
using regrassion
RF Selection

4. Patient history 672 variables scattered into 20
categories as the input to model
Dual-energy x-ray absorptiometry (DXA) scan
results on bone mineral density (BMD) variation
on different visit
Bone loss rate
Data Set

5. Typical Risk Factors

6. Informative Risk Factors Generated
These factors are used to determine
bone loss

7. ROC (receiver operator
characteristic )
PR (precision-recall )
RBM (Restricted
Boltzmann Machine)
FT(fine-tuning)
AUC (area under curve )
Performance Study for Osteoporosis
Prediction
Error rate
1.0 indicates a perfect performance

8. Osteoporosis Prediction Based on
Informative RFs

9. Recall and Precision

10. SOF(Study of osteoporotic fracture )
SOF data Set statistics

11. Thanks!