Google Lens Image Search Now Able to Identify Skin Conditions

2023-06-16

"Google Lens now provides more information about annoying rashes that you are unsure whether to worry about. In a blog post released this week, Google outlined how its built-in Lens image search feature in its iOS and Android applications can 'search for skin conditions' such as 'strange moles or rashes'. It can also be used for other parts of the body if you want to learn more about lip protrusions, nail patterns, or scalp hair loss. The blog post states, 'Simply take a photo or upload a picture with Lens, and you'll find visual matching results to provide search information.' However, Google specifically warns that the results are for reference only and not a diagnosis, and advises users to 'consult a medical professional'. Google has been exploring the use of artificial intelligence image recognition technology to identify skin issues for years. At the 2021 I/O developer conference, Google showcased a tool that attempts to identify skin, hair, and nail problems using photos and survey responses. At the time, Google stated that the tool can identify 288 different conditions and provide the correct diagnosis among the top three suggestions in 84% of cases. As of now, Google's website for its DermAssist tool shows that it is 'undergoing further market testing through limited release'. The website also mentions that while the tool has obtained the CE mark as a class I medical device for the European Economic Area, it has not been evaluated by the US Food and Drug Administration and is 'for reference purposes only, not for medical diagnosis'. While it is indeed feasible, this does not prevent people from attempting to use such tools for diagnosis. It can be said that adding such disclaimers simply shifts the responsibility to the users while still allowing Google to provide the same basic service. There are valid reasons to exercise caution with AI diagnostic tools. In terms of identifying skin issues, a persistent criticism is the lower accuracy of such software on individuals with darker skin tones. A study cited by The Guardian in 2021 pointed out the lack of skin type category data in many publicly available image databases used for training AI systems, and the absence of images of individuals with darker skin in databases that do contain such information. Google has been working to address these issues. Last year, Google collaborated with Harvard professor Ellis Monk to promote his Monk Skin Tone (MST) and best practices for using it in AI development. Google stated that its team has been using the MST for their own 'machine learning annotation tasks, ethnographic research, and fairness testing'. In 2021, Google also claimed that its deep learning system is actually more accurate in identifying skin conditions in black patients. Slides provided by Google to Motherboard showed that the company's system achieved an accuracy rate of 87.9% for black patients, higher than other races. Source: https://www.theverge.com/2023/6/15/23761905/google-lens-skin-conditions-rash-identification"

RECENT AI NEWS

RECENT AI TOOLS