Dermatology (skin diseases) is well suited using Artificial Intelligence (AI) research

Artificial Intelligence (AI) studies in radiology for both mammography[1] and pulmonary CT scans[2] demonstrated that radiologists’ performance, especially specificity, increased with the use of an interface desktop AI tool to interpret the images.

Try our FREE skin image search and get the right answer now

or ask a real dermatologist

Ask our board certified dermatologists today

Convolutional Neural Network (CNN) limitations

The limitations for a good working Artificial Intelligence (AI) model using Convolutional Neural Network (CNN) model are the quality of images per classification together with a limited data set per classification. The Thrun laboratory at Stanford USA that published a letter in Nature, used a dataset of 129,450 clinical images (clinical images were high resolution in focus images, many times taken by a professional medical photographer) consisting of 2,032 different diseases[3]. It was tested on two classes: keratinocyte carcinomas versus benign seborrheic keratoses; and malignant melanomas versus benign nevi. The CNN was on par at diagnosing correctly compared to 21 dermatologists on the two classes.

First Derm – automated dermatology answers

First Derm has over the years built a database of many 100s of thousands anonymous smartphone images. The database includes images of various quality of about 250 different skin conditions. During 2017 we applied the image set to a CNN with the goal to identify skin lesions from skin concerns taken with a smartphone.

Try our FREE skin image search and get the right answer now

Preliminary results

Version 0.1 of our web interface is trained on 33 skin disease classes (inflammatory, moles and STDs). The accuracy average is 40% on one skin disease and 80% for top 5 skin diseases. *

Version 1.0 of our web interface is trained on 43 skin disease classes (inflammatory, moles and STDs). The accuracy is 37.8% on one skin disease and 73.2% for top 5 skin diseases. *

Version 2.0 of our web interface is trained on 44 skin disease classes (inflammatory, moles and STDs). The accuracy is 49.3%% on one skin disease and 81.7% for top 5 skin diseases. We are continuously experimenting to make it better. *

Version 3.0 of our web interface is trained on 68 skin disease classes (inflammatory, moles, STDs and Covid-19 related skin diseases). The accuracy is 31.7% on one skin disease and 68.1% for top 5 skin diseases. *

*The above results are tested with 20% of a data set that has not been trained. We also test the models manually and the tendency is that for skin diseases that have more images represented the accuracy is higher.

We have an open API that can integrate into any internet platform and apps. Currently health chat bots and telemedicine primary care clinic use the API. If interested you can sign up here

Ask our board certified dermatologists today

Here are some more examples you can read about

ICD 10 L70.0 ICD 10 L70.0 Acne help with artificial intelligence (AI)
ICD 10 L30.9 Dermatitis (eczema)
ICD 10 L73.9 Folliculitis (hair root inflammation)
ICD 10 D22.0 Nevus (mole)
ICD 10 L82.9 Seborrheic keratosis (senile mole)
ICD 10 A63.0 Human papillomavirus – HPV (genital warts)
ICD 10 N48.1 Balanitis (male genital eczema)

References

[1] Intelligent CAD workstation for breast imaging using similarity to known lesions and multiple visual prompt aids. Giger M et al. Proc. SPIE 4684, Medical Imaging 2002.
[2] Machine Learning and Radiology. Wang S, Summers RM. Medical Image Analysis. 2012;16(5):933-951.
[3] Dermatologist-level classification of skin cancer with deep neural networks. Esteva A, Kuprel B et al  Nature 017/02/02/print

 

Try our FREE skin image search and get the right answer now