PracticalDermatology

The search giant Google may be trying to give us a run for our money with their new artificial intelligence (AI)-powered dermatology assistance tool.

Previewed at Google’s I/O developer conference, the new tool promises to identify dermatologic issues using a smartphone camera. This product has been CE marked as a Class I medical device in the European Union but is not yet available here.

Here’s the rationale behind its creation: There are almost 10 billion Google searches related to skin, nail, and hair issues each year, and there is also a shortage of dermatologists in many rural areas.

To the rescue: The new web-based dermatology app.

A user launches the tool, takes three images of the skin, hair, or nail concern from different angles with the phone’s built-in camera, answers some rote questions about skin type and other symptoms, and voila: the model quickly analyzes this information, draws from its knowledge of 288 conditions, and spits back a list of possible diagnoses.

For each match, the tool will also present dermatologist-reviewed information and answers to commonly asked questions about the condition, along with similar matching images from the web.

It isn’t intended to provide a differential diagnosis or be a substitute for seeing your dermatologist for an in-person examination, and additional testing like a biopsy may be needed. If accurate (and that’s a big if), this tool could be useful for self-triaging, and it could also help non-dermatologists decide whether and when to refer a patient to a specialist. Google is also reportedly working with a Stanford University research team to test how well the new tool works in a health care setting.

Google has published several papers that validate their AI model including a piece in Nature Medicine that explains the deep learning approach and showed that Google’s AI system can achieve accuracy that is on par with that of dermatologists.¹ The study pitted dermatologists against AI to diagnose keratinocyte carcinomas versus benign seborrheic keratoses or malignant melanomas versus benign nevi; the deep convolutional neural networks detected skin cancer as accurately as 21 seasoned dermatologists. A more recent paper in JAMA Network Open showed that non-dermatologists can use such AI-based tools to improve their ability to diagnose skin conditions.²

The tool comprises de-identified data from around 65,000 images and case data of diagnosed skin conditions, millions of curated skin concern images, and thousands of examples of healthy skin across all different demographics.

Google says that their model accounts for age, sex, race, skin types, and tones. According to media reports, 90 percent of the database was made up of people with fair skin, darker white skin, or light brown skin. Some dermatologists have pointed out that this data shouldn’t be trusted to diagnose skin of color. In response to such criticism, Google told Reuters that it is developing an alternative to the industry-standard method for classifying skin tones.³

My Take

This app—maybe with some tweaks—will be here soon, and our patients, as well as non-dermatologists, may well use it.

There are risks. The AI-generated responses could foster a false sense of security or unnecessary anxiety. Poor lighting or focus could mar an image, resulting in a misdiagnosis, which could forestall specialty care, and that is a major concern.

The bottom line? Google has reach, so it’s quite possible that when you ask patients what brings them in, they may say “Dr. Google sent me” even more so than they already do.

1. Liu, Y., Jain, A., Eng, C. et al. A deep learning system for differential diagnosis of skin diseases. Nat Med 26, 900–908 (2020). https://doi.org/10.1038/s41591-020-0842-3

2. Jain A, Way D, Gupta V, et al. Development and Assessment of an Artificial Intelligence–Based Tool for Skin Condition Diagnosis by Primary Care Physicians and Nurse Practitioners in Teledermatology Practices. JAMA Netw Open. 2021;4(4):e217249. doi:10.1001/jamanetworkopen.2021.7249

3. Paresh D. “EXCLUSIVE Google searches for new measure of skin tones to curb bias in products.”

https://www.reuters.com/business/sustainable-business/exclusive-google-searches-new-measure-skin-tones-curb-bias-products-2021-06-18/