Document – Model Dermatology

This is an EXPERIMENTAL DEMO APP for Medical Professionals.

This program will predict skin disease from the image findings.

The results obtained with this program are for research purpose only.

Please check with your dermatologist for accurate diagnosis.

** THIS DEMO CANNOT CONFIRM THAT THE SUBMITTED IMAGE IS NOT SKIN CANCER. Approximately 10% of skin cancer can be missed if diagnosis is based on clinical photography alone.

** The photographs used for analysis are transferred, and stored and can be used as research data for new service research and development.

** This DEMO produce a sub-optimal performance for the images of non-medical professionals. Please be sure to take pictures in bright places.

– Malignancy Prediction

The app will display the result text according to the following conditions:

“Low” : miss rate – 0 ~ 10%
“Medium”, Biopsy recommended : miss rate – 10 ~ 20%
“High”, Biopsy recommended : miss rate – 20 ~ 50%
“Very High”, Biopsy recommended : miss rate – 50 ~ 100%

cf) miss rate (false negative rate) = 1 – sensitivity; The two scores (output and index) in parentheses must satisfy the condition at the same time.

– Raw outputs

q – image quality output
n – image nodularity output
M – malignancy output, index
S – steroid output, index
A – antibiotics output,index
V – antivirals output,index
F – antifungals output,index

– Accuracy
The model has approximately 88% accuracy (sensitivity and specificity) for malignancy detection.
The Top-1, 3, and 5 accuracy of this algorithm are around 45%, 70%, 80%.