Internet search on skin diseases compared to Skin Image Search™
80% of people search their health symptoms on the internet and try to self-diagnose (PEW study done in 2012). A lot of people do not realize this, but many of the images on the internet have been labeled incorrectly, so even if you find an image that looks similar to your skin condition, it can have been labeled wrong.
Try our FREE Skin Image Search today and get peace of mind
Internet skin disease search
We searched for Actinic Keratosis using Google and we clicked the image tab to find relevant images.
Example of an internet search for Actinic Keratosis
We then clicked on one of the images. A larger image pops-up with a respective link to the website and in the right corner there are similar/ related images.
Clicked on the image and is said Actinic Keratosis. To the right there is a display of similar/ related findings.
We then clicked to the related website, Quality Dermatology – you then come to information related to Actinic Keratosis and the image is embedded in the informational text.
The image is a Seborrheic Keratosis, a senile wart, which was confirmed by a board certified dermatologist.
We then submitted the same image via Skin Image Search™
Results: Internet lesion that was labeled as Actinic Keratosis. The Skin Image Search™ thinks it is a Seborrheic Keratosis. A board certified dermatologist confirms the image to be a Seborrheic Keratosis (senile wart).
What is Skin Image Search™ and can it be trusted?
Skin Image Search™ algorithm has been trained on hundreds of thousands of smartphone images from First Derm app users. Over 250 different skin diseases have been submitted over the years. The majority of the queries have been on about 40 skin diseases, which is 80% of First Derm users. The images have been used to train a deep convoluted neural network (CNN).
Using skin images to train a CNN has been studied before, specifically on pigmented lesions such as moles and melanomas, where the results have been similar to that of dermatologists. Basically a CNN is given images and data on a much grander scale in a shorter time frame than what a dermatologist comes across in their entire life. So of course with the correct data and correct training, a CNN will be better than a human on identifying skin diseases, but it will take some with time to get there.
Skin Image Search™ current algorithm has been trained on 33 different skin diseases, it is 40% correct on one skin disease class and 80% correct on the top five skin disease classes (differential skin diseases).
Skin Image Search™ is currently not as good as a dermatologist, but better than many non-dermatologist physicians. It should be used as an informational guide on skin diseases and it does not replace a dermatologist. However it is much better than just a random internet search to help guide you to relevant information on your skin condition. You can read a testimonial from a user here
Try it our here Skin Image Search™ or you can ask a real dermatologist
The Specialist doctor from the University Hospital in Gothenburg, alumnus UC Berkeley. My doctoral dissertation is about Digital Health and I have published 5 scientific articles in teledermatology and artificial intelligence and others.