Nowadays, it seems everyone is talking about AI, yet we may wonder if this new technology is good or bad for society. Before making a judgment on AI, it’s important to understand what AI is, especially if you’re a dermatologist. I recently helped write a review in Dermatological Surgery looking at the use of AI in dermatology, including how it can be used for skin cancer testing and app development for excisions and surgical procedures. I hope this article can help you better understand what AI can potentially mean for dermatologists like you.
AI is a broad term that covers the use of machines to stimulate human cognition to solve problems. It’s important to understand the subsets of AI, including machine learning, which enables systems to learn from patterns without being programmed. An even further subset is deep learning, which will be more applicable in dermatology. Deep learning employs artificial neural networks (ANNs) inspired by the human brain. ANNs, when programmed correctly, can recognize patterns and make predictions using these “neurons” based on data and previous encounters.
So what does all of this mean for the practicing dermatologist? AI is going to change the way we diagnose and treat – for the better.
Still skeptical? Stay with me. Here’s my vision for AI in dermatology:
Image Analysis and Diagnostics
Skin cancer detection is a big part of what we do as dermatologists, and it’s important to identify dangerous conditions, such as melanomas, as early as possible. We can use AI algorithms to detect and predict skin cancer. If trained well, these algorithms can differentiate between benign and malignant lesions. Dermoscopy, perhaps one of the more difficult skills to learn and master, can be enhanced with AI automation leading to earlier detection of skin cancers.
Prediction
Medicine relies on prediction, and we commonly make risk assessments. Deep learning can use data to help us determine the likelihood of an eventual diagnosis of melanoma or other skin conditions based on genetics, lifestyle factors and medical history. AI can also predict the progression of a lot of skin conditions, which provides more insight for both the physician and the patient.
Guiding Therapeutics
With the development of new dermatological therapies seemingly every day, there is a vast opportunity for AI to influence these therapies and their clinical utility. Psoriasis is one of those conditions that affects a decent portion of the population where there is no guaranteed way to predict which medication will work for each patient. Therefore, we throw everything at the wall and see what sticks. Yet it takes time to find a treatment that works, and we don’t want to inject patients with medications that provide no benefit. Patients want to see quick improvement and they don’t want to waste time. The current process is also costly for insurance companies and patients. Now there’s a movement to use AI to predict which patients will respond best to certain treatments based on their skin microbiome. Once patients are on an effective medication, automation takes out the conjecture in the process. We already have the data but translating that into prediction is where AI comes in.
Optimizing Medication Use
What happens now in dermatology is that you get handed a laundry list of meds when you come in for an initial visit and no one really understands the interaction between drugs beyond surface level. AI can help us to better understand the interplay between these medications while also providing an optimal concoction of the most effective, personalized plan.
One way I think AI can help is with patch testing. Currently, we test patients using common allergens yet often we don’t get a clear answer. I envision AI to use morphology and anatomical location as data points to highlight the more likely allergens responsible for the patient’s rash. For example, if I see a specific rash in a particular distribution, I am more likely to think about a clothing or textile allergy.
Research and Development
AI has potential for drug discoveries, such as predicting how different compounds interact with skin cells and with each other. AI also can ensure compounds are safe and effective before public distribution. One of the major hurdles in clinical trials is the time required for each phase. AI can streamline the process, identify study patients and monitor side effects, all in real time.
Education
AI is prone for the education space as it can power simulations and supplement training in various areas including skin cancer surgeries, complex procedures, such as hair transplants, and lesion detection in skin of color populations.
Of course, we have to be cautious with any new developments. In a way, it’s good that healthcare is often late to the party with its restrictions and red tape. We can learn from the discussions currently happening in the tech sector over safety and privacy concerns. The provisions required in the financial and technology fields can be a template for how we incorporate AI in healthcare. Mistakes can help us establish quality guidelines for AI use in healthcare.
AI has a significant potential in improving the diagnosis, treatment and management of dermatologic diseases. It’s here to stay and I hope as dermatologists – and patients – we can welcome it because the faster we do, the more effective we become for our patients and ourselves.
