Artificial Intelligence Melanoma (Skin Cancer) Detection based on Deep Learning
- Model Melanoma -
This is an EXPERIMENTAL DEMO APP for Medical Professionals.
PLEASE BE SURE TO TAKE PICTURES IN BRIGHT PLACES.
This is an artificial intelligence skin cancer (melanoma) detection program based on Deep Learning Algorithm.
The results obtained from this program should be used for reference purposes only. Please consult your dermatologist to get a confirmative diagnosis.
If the result is a melanoma, it is highly recommended to visit dermatology.
However, because melanoma cannot be definitely diagnosed by clinical photography alone, therefore, even if the result is not a melanoma,
THIS APP CANNOT CONFIRM THAT IT IS NOT A MELANOMA.
Approximately 10% of melanoma may be missed when we diagnose with clinical photography alone.
The photographs used for analysis are transferred, and stored as a database, which will furnish a reference framework for a future research.
The app will display the result text according to the following conditions:
1) "Probably Not" : miss rate - 0 ~ 6.5%
2) "Possibly", Biopsy recommended : miss rate - 6.5 ~ 19%
3) "Likely", Biopsy recommended : miss rate - 19 ~ 50%
4) "Very Likely", Biopsy recommended : miss rate - 50 ~ 100%
cf) Miss rate (false negative rate) = 1 - sensitivity; The two scores in parentheses must satisfy the condition at the same time.; We calculated the sensitivity of the model with Edinburgh Dermofit Image Library.
- Accuracy of the Model
The model has approximately 90% accuracy. The AUC value for melanoma detection with Edinburgh's melanoma and nevus (Caucasian dataset) is up to 0.974, which is currently the most outstanding result.
The AUC value for melanoma with Asian dataset is up to 0.961.
- Papers (http://www.modelderm.com)
Classification of the clinical images for benign and malignant cutaneous tumors using a deep learning algorithm - http://www.jidonline.org/article/S0022-202X(18)30111-8/fulltext
Interpretation of the Outputs of Deep Learning Model trained with Skin Cancer Dataset - https://www.jidonline.org/article/S0022-202X(18)31992-4/fulltext
- iPhone User (Safari)
http://melanoma.modelderm.com
- Model Dermatology (AI Skin Disease Diagnosis)
https://play.google.com/store/apps/details?id=com.phonegap.dermatology_en
http://modelderm.com