fbpx
Connect with us

Opinion

Are smartphone apps the key to spotting skin cancer?

Published

on

With increasing numbers of smartphone applications being created to assess the risk of someone developing skin cancer, Dr Iram Ashraf, dermatology tutor at online learning provider Learna, explores the benefits and pitfalls of an increasing reliance on technology.

Since 2010, there has been a tremendous surge in the number of mobile phone applications which are designed to help identify potential melanoma or non-melanoma skin cancer.

They mainly offer two types of service – offering users guidance on whether they should seek a medical consultation for a skin lesion or anomaly they have photographed with their device, and also a way to record and monitor lesions over time to see if any changes happen.

While these apps started off as simple tools to analyse patient-generated images of lesions next to ones stored in online libraries, recent years have seen more smartphone apps converted into machine-learning artificial intelligence programmes, working off big data using convolutional neural networks.

In my experience, some of these apps are of sufficient quality so that it is possible for both doctors and patients can benefit from them.

Localising lesions is an important step in identifying the potential development of skin cancer, providing the opportunity for medical professionals to then provide health education to patients using these apps.

The priority for the medical community should be early diagnosis and prompt treatment.

We are achieving this on the whole, but there is of course ample room for improvement, and the more effective tools that are available to the public to help expediate the process further, the better outcomes there will be.

Fallibilities

But, of course, both AI and the professional naked eye are far from infallible.

Histological confirmation is still the gold standard, but when we read the literature and systematic reviews we find that 74-84 per cent is the rate at which we as practitioners can confirm any disease.

And Melanoma is such a malignant condition that we can sometimes have just a vague idea of its nature and this can lead us to miss problems, which can obviously be hazardous for a patient.

However, even so, as doctors with a greater degree of individual context for each patient than any app could ever hope to achieve.

For example, the determinations from these apps are at the mercy of all sorts of crucial and contextual factors related to where an image is captured, by who, and under what conditions.

Dr Iram Ashraf

For example, if we are comparing an image of a lesion on a person with light skin, we are not comparing like-for-like with an image of a lesion on a person with dark skin.

Similarly, a person living in Iceland has a different risk level for developing melanoma, perhaps due to sun exposure, when compared to someone living in Pakistan.

Additionally, many apps that we have seen cannot identify the proper age of a person, which is an important factor when assessing the risk that skin cancer poses to a patient.

Children, for instance, can be considered by themselves or their parents as having some malignant skin lesion when in actuality there is nothing to worry about.

Therefore, if AI is drawing from information gathered across the world from entirely different types of people, it is likely to produce inconsistent and unreliable results.

If there is a high risk of bias, then, at best, there will be unnecessarily high levels of referrals which will increase workloads for already busy dermatologists and, at worst, people who are actually at risk from melanoma will wrongly put their minds at ease and not seek consultations due to unreliable methods of analysis.

Balance, and working together

And so there must be a balance in reliance on this evolving medical digital technology.

However, this balance will potentially be problematic to achieve in the years to come as AI and algorithms continue to disrupt global healthcare and will continue to do so at an increasing rate, not just because the technology will improve, but to fill the gaps left by a shortage of healthcare workers across the world.

Most apps are, at least initially, not built with the help of dermatologists, nor set up so that results can be monitored by them.

In order to give the public tools that can truly be relied upon to identify skin issues, their creators need to seek greater active contribution from dermatologists and the medical establishment to see and identify the lesions.

AI can do so much, but real human medical expertise is needed at the outset (at the very least) in order to set the parameters of where it draws information from. Without this AI provides no real intelligence at all.

Increased partnership between app developers and dermatologists will hopefully lead to a more reliable filtering of patients, aiding healthcare workers at a time when their services have never been under more pressure.

Dr Iram Ashraf is a tutor in Dermatology at online learning provider Learna, and has 13 years of experience practicing at The Concept Aesthetic Skin Clinic, Lahore, Pakistan.

She is an MBBS, DipDerm (UK), European board certified dermatologist (UEMS-EBDVD), holds an MSc Aesthetic Medicine Queen Mary University of London, international trainer mesotherapy, DGM (German Society of Mesotherapy) and is currently based in Germany.

Trending stories