09. January 2025
Dermatology in the era of Artificial Intelligence
An interview with Prof. Dr. Holger Hänßle,
University of Heidelberg
Medicine and artificial intelligence (AI) are becoming increasingly intertwined, also in the field of dermatology. Prof. Dr. Holger Hänßle has significantly shaped this innovative field of research and, together with his team in Heidelberg, has published numerous groundbreaking studies. In our interview, he discusses his career, his experiences with FotoFinder, the potential of AI and the changes in dermato-oncology.
FF Prof. Dr. Hänßle, how did you get into dermoscopy?
Prof. Dr. Holger Hänßle I've always been visually inclined, with a strong interest in morphology recognition. During my studies, I was torn between radiology and dermatology. However, the limited patient interaction in radiology led me to choose dermatology. My passion for computers and technology naturally drew me to the video dermatoscope—a device that allowed me to immerse myself in my technical interests as well.
FF How long have you been researching in the field of AI? How did you become interested in this topic?
Prof. Dr. Holger Hänßle I worked and published in dermoscopy for many years, particularly in the field of sequential digital dermoscopy. During that time, we were already using a FotoFinder with the Moleanalyzer. When Andre Esteva's study from Stanford was published in Nature in 2017, it became evident to me that artificial intelligence had reached the level of well-trained dermatologists. We fully immersed ourselves in this field, and in 2018, with our first publication**, we became the second to publish on this topic. We were able to confirm the Stanford results and gained significant attention by showing that these results could be reproduced using a completely different algorithm from a completely different country.
* “Dermatologist-level classification of skin cancer with deep neural networks”, Esteva et al., 2017
** “Man against machine”, Annuals of Oncology
FF When and how did you first come across FotoFinder?
Prof. Dr. Holger Hänßle That was quite early in my career, when I was a young assistant doctor at the University of Göttingen. We didn't have a proper device at the time, so I wrote a letter to Mr. Mayer Senior, asking for a system for our scientific work. Remarkably, we received a loaner device, and from that moment on, I have always felt a strong connection to FotoFinder´.
FF What impresses you most about AI?
Prof. Dr. Holger Hänßle What impresses me most is that AI never forgets. In my lectures, I always ask: how many melanomas have you seen, or how many can a dermatologist expect to see in a lifetime? The number is usually limited to a few thousand. Now, consider that the FotoFinder AI has been trained on over 180,000 images, all of which it retains perfectly. This is the significant advantage – it's why an algorithm like this can ultimately outperform a doctor.
FF Who would you recommend AI to? Even experienced dermoscopy experts?
Prof. Dr. Holger Hänßle Yes, even dermoscopy experts, because various studies have clearly shown that even top specialists can improve their performance with AI. However, it’s the beginners and doctors with average experience who benefit the most. Even as an expert, I sometimes feel uncertain. In these moments, I use AI for a second opinion and gain an additional perspective, which I then incorporate into my diagnosis.
FF What key findings from your studies would you like to share?
Prof. Dr. Holger Hänßle I think it's crucial for users to understand that AI is an assistance system, like a lane departure warning in a car. You can have multiple assistance systems that increase safety and help prevent serious oversights. However, we're not at the stage where AI can be considered autonomous driving. AI has been trained on specific data, evaluates changes based on that training and provides alerts, similar to the lane assistant’s warning, 'Watch out, there might be something!' But it's just one part of the equation.
The patient's age, medical history, risk factors—these elements are equally important. People often seek the convenience of having a machine make decisions for them, but that's where the danger lies, as it can lead to significant errors. The doctor remains in control, steering the vehicle according to their judgment. That's very important to me. Perhaps the most crucial advice is not to trust AI blindly.
FF What is your conclusion on workflow and practice, especially regarding the interaction between Total Body Mapping, dermoscopy and AI.
Prof. Dr. Holger Hänßle I believe FotoFinder is one of the companies that places great emphasis on workflow, which is why user ergonomics are excellent. Everything integrates seamlessly—the cross-connections between overview images and dermoscopy are particularly effective. Of course, each practice needs to establish its own workflow: who takes the images, what tasks can be delegated to the staff and what responsibilities remain with the doctor. For me, the interaction between workflow and the image quality provided by FotoFinder is crucial.
FF What do you think skin checks will look like in the future?
Prof. Dr. Holger Hänßle I envision we would use Total Body Mapping to examine the patient and be able to say, even during the screening, what is very likely benign. This would allow the doctor's attention to focus on just two or three skin lesions. The software would use a red flag to indicate where to look, which would really make screening easier. At the moment, the doctor still thoroughly examines the patient themselves. But I believe that the resolution of full-body images will continue to improve, enabling AI to be used in this area as well.
Three personal questions at the end
For me, dermoscopy means:
Passion
A tip for your younger self:
Be more patient
Life motto:
The joy of achievement