Artificial Intelligence in Dermatology Is it just the start?
Rudrakshi Shetty
Final year MBBS, Northern State Medical University, Russia
Artificial Intelligence (AI) as a concept is revolutionary and daunting at the same time. On one side, we see the endless possibilities we can achieve through AI but on the other end, we’re left with the question, where does it all end? Is its function restricted to assisting humans or can it completely take over? It has the ability to replicate human intellect and learn from experiences. AI is supposed to play an “imitation game” where just like humans, it can learn from mistakes, self-train and apply to new situations. Except, it can do it much more efficiently than humans. The fuss around the topic might seem recent but the use of AI has been around for quite some time now, even in the field of healthcare.
The use of AI in the field of medicine has been used to assist in accurate diagnosis, therapy and prognosis. The work of AI can be easily confused with statistical differences but the core difference between the two is that AI uses pattern recognition and data mining to analyze structured and unstructured data. AI was first used in healthcare in 1976 to diagnose acute abdominal pain. This revolutionary move was made by Gunn with the help of computer analysis.
Other than the general use of AI in medicine, it started to pave its way into specific fields such as psychology, dermatology and radiology. With regards to dermatology, the magic of AI was out to use for the diagnosis of Malignant Melanoma. Wilhelm Stolz and the team in Munich developed a handheld dermatoscope for cutaneous microscopy. This step led to more enhanced ways of detecting lesions using computer-aided diagnosis. This usually involved observing the lesion through either photography, spectrophotometry and/or dermoscopy. Many softwares employ C programming to accurately analyse lesional borders, changes over time and any malignant features fed into the system. Malignant Melanoma was just the stepping stone, AI has been beneficial in diagnosing psoriasis, acne and rosacea, autoimmune disorders (in terms of classification, risk prediction, disease progression, monitoring and more), allergic contact dermatitis (measures genomic biomarkers using Genomic Allergen Rapid Detection Assay) and ulcer evaluation to name a few.
While the wonderful aspects of AI can be endless, as something that’s still under experimentation, it is even more important for us to focus on its limitations. For instance, AI can only ever detect what is already present in its database. If we were to take into account the complexities across various institutions, regions and races, AI might not seem as appealing. For example, when talking about Lyme disease, AI may be able to detect the characteristic ‘bull’s eye’ lesion but if it were to diagnose the same in a patient of colour like a black patient, the diagnosis may be excessively delayed or absent altogether since they don’t show the characteristic lesion.
Accepting care simply based on algorithms can be difficult for patients. In dermatology, biopsy remains the gold standard for the most part which can be assisted by AI. We also face the ‘Black Box Problem’. Usually, for any diagnosis, a doctor will have a particular reasoning for the same but when it comes to AI, it uses information bulk within hidden layers. Among all these, determining the exact cause of why a particular diagnosis has been made by AI can be challenging. Healthcare doesn’t simply mean diagnosis it extends even beyond treatment. Considering this, the scope of AI seems particularly restricted and we may experience resistance from patients if we were to completely depend on it. Among all these, the breaching of patient privacy remains the biggest concern.
With the introduction of every innovation, it is up to us to use it for the best of our patients and healthcare workers. Using AI as an adjunct to our current system can help build our infrastructure and lead to a more productive functioning rather than relying on it completely. Opinions might vary, and many people might agree on varying degrees of involvement of AI but if we keep the best interest of our patients at the centre of decision-making, the use of AI can turn into a blessing after all!
