28 June 2023

What’s the similarity between John Connor and a dermatologist?

Clinical Technology

AI is coming for your dermatology practice.

Dermatologists need to be actively involved in the development, evaluation and implementation of AI tools, an Australian expert has warned.

The price of not being part of the technological disruption that will trigger new clinical patient workflows?


“John Connor is the leader of the worldwide human resistance against the machine, [but dermatologists] lead the human intelligence to ensure the safe, effective and ethical translation of AI [in practice],” said Dr Tony Caccetta, a Perth-based dermatologist, told delegates at the Australasian College of Dermatologists’ annual scientific in Sydney last month.

Dr Caccetta pointed to three ways AI can improve dermatology – provided it is done in a regulated manner, as there is currently a lack of strong evidence supporting the safe and effective use of AI in dermatology.

Rapid process automation

Dermatologists have not been spared from the workforce shortages affecting the broader healthcare industry. Consequently, it is very important to try and prioritise patients seen in practices.

The problem with this, according to Dr Caccetta, was this kind of work is repetitive and time consuming for both administrative and clinical staff, with processing each patient from start to finish taking more than 10 minutes.

The first area where AI and automation may assist practices is rapid process automation, where AI can interpret referrals and image data, create bookings based on urgency, subspecialisations or interests and insert all relevant materials into the electronic medical record.

Dr Caccetta is currently developing rapid process automation software called DermTriageAssist, which will undergo real-world evaluation at his practice, Perth Dermatology Clinic.

“We hope to demonstrate, on the smaller scale, the time saving potential of this software and its human equivalent performance. There are similar larger scale applications undergoing evaluations in the NHS in the UK,” Dr Caccetta explained.

Image recognition

A more familiar use of AI in dermatology is image recognition, where AI is applied to images to provide information, recommendations or risk ratings to either patients or healthcare professionals. This use of AI creates a new workflow, as it has the potential to improve patient access and reduce healthcare costs.

Dr Caccetta pointed to Google’s DermAssist as an example of how image recognition can be used in dermatology. Patients upload three photos of their skin condition and answer a few questions before DermAssist provides them with a list of possible matching skin conditions and information about each of them.

DermAssist can identify almost 300 different skin, hair or nail conditions after being built in collaboration with dermatologists and being trained on millions of images.

Large language models

Large language models such as ChatGPT and Google Bard have received a lot of attention in recent months. These models involve training AI on text or voice data rather than images alone.

Dr Caccetta explained that large language models have “infinite potential applications” in the real world as they can serve as virtual clinical or administrative assistants. For example, an AI chatbot could be used as a first point of contact for patients seeking additional information about dermatological conditions.

Alternatively, ChatGPT can be used to write letters to referring doctors based on consultation notes input by a dermatologist. After playing around with this himself, Dr Caccetta has been impressed with the “significant amount of empathy and creativity” included in the finished product that was not present in the initial notes.