The MelanomeFrance forum is an online private forum of around 1500 Melanoma patients and carers, mostly from France, but also French speaking Belgians, Swiss, North Africans and Canadians.
Melanoma patients have a variable understanding and knowledge of Artificial Intelligence in their Cancer care. They are generally aware that the use of Machine learning is increasing in Dermatology and Oncology. We have discussed some of these issues on our Patient Forums. The areas where Melanoma patients are most interested in the use of AI, are better diagnosis, risk of recurrence prediction and treatment decision making.
Diagnosis
Patients are aware that some melanoma subtypes and some lesions are harder to diagnose – spitz, desmoplastic or regressed primary. Awareness of the need for more effective identification of melanoma primaries stems from experience of misdiagnosis in primary care, requests for second opinions and presence of familial risk.
Patients are aware that more than 50% of suspect lesion referrals from primary care to specialists are patient-instigated. Patients often feel that annual skin checks are not frequent enough and will choose to use the skin monitoring apps between dermatology visits for reassurance. Patients frequently share photos of their lesions on the closed patient forums, now that mobile phone technology allows this. Though the response is to always seek expert advice since biopsy is the only certain solution for now, we have found that during the restrictions of the Covid19 pandemic and because of a serious long-term shortage of Dermatology professionals in France, people at risk and patients, have been using skin monitoring phone applications more frequently. Previously, patients were using these skin lesion identification apps (eg Molemapper, Myskin, Skinvision etc.) with limited levels of trust, largely because health professionals have not been encouraging their use until they become more reliable or useful. Their main advantage has been to take sufficiently good photos to share with their dermatology HCP and to conveniently self-monitor or check family members. While most patients are unaware of how the apps are trained or how the risk algorithms and tools are developed, their reliability is often discussed on Forums.
Some patients have annual access to hospital or dermatology practice based whole-body photography and image analysis, but this access is very patchy across the country. More anxious patients will seek this service out where possible. The most frequently used in France is the Fotofinder system and the Molemax in Belgium – both image analysis and Machine learning risk assessment systems. In this context the tools developed into the iTOBOS project raised the interest of patients, especially of those considered at high risk of getting melanoma.
Patients are also aware that Machine learning algorithms can match or exceed that of human dermatology expertise but are unsure whether these are widely used outside of research, though many patients think access to this would reduce their anxiety and improve their outcomes. One major issue is the lack of confidence of the healthcare practitioner that these devices could reduce the burden on themselves and maintain optimal diagnosis. Thus education is necessary for both patients and physicians.
Risk Prediction
Patients are aware of Familial Risk even though this involves a relatively small number of patients, and they are often aware of other risk factors that increase their chance of more primaries and/or progression. A few patients are sent for genetic testing but we often have to help patients understand that this only supports preventive behaviour and adherence to self-checking and diligent follow-up. We strongly focus on educating patents and carers in the importance of understanding staging and its relationship to risk. Discussions are had on tools that analyse phenotype, lifestyle, family history and characteristics of the primary to estimate the chance of lymph node progression and metastasis, and patients have some knowledge of the place of machine learning algorithms in this prediction. They frequently ask if primaries can be tested for risk of spread beyond just the staging characteristics – to combine better primary imaging, polygenic risk, strip-tests, more detailed histopathology and molecular assays. They are frustrated by slow validation processes.
Treatment decisions
Melanoma patients are seeing their therapies moving into earlier stages, but with huge uncertainty in overall risk and benefit, they are expected to make decisions with very little personally relevant information. Many patients know that inputting relevant tumour/genetic/phenotypic/clinical data into AI algorithms could be helpful in determining whether a patient needs earlier treatment or if they can avoid the risk of long-term toxicity or un-necessary surgery. For now, a lack of credible data, un-assuaged fear of recurrence and slow development of effective decision tools, are leading to overtreatment, which leaves many patients with chronic toxicity or sub-optimal treatment. AI is just beginning to show encouraging results in combining polygenic risk, primary tumour analysis, genome sequencing and molecular-level criteria predisposing to toxicity. Effective tumour and patient avatar modelling may provide some answers to how to design Machine learning algorithms to prioritize information that helps treatment choices. Patients are aware of molecular diagnostic assays on the horizon, eg Decision DX, Skyline DX etc that may help avoidance of surgery or un-necessary treatment.
For now, French patients look forward to the advances in validation and implementation which will permit them to develop trust in these processes. We still have challenges for patients and clinicians to be satisfied that machine learning leads to secure data intelligence and use, a better understanding of tumours and their environment, more precise treatment strategies and better access to effective decision-making processes.
Gilliosa SPURRIER-BERNARD
President & Founder MélanomeFrance
Core Member Melanoma Patient Network Europe
References
-
Avilés-Izquierdo JA, Molina-López I, Rodríguez-Lomba E, Marquez-Rodas I, Suarez-Fernandez R, Lazaro-Ochaita P. Who detects melanoma? Impact of detection patterns on characteristics and prognosis of patients with melanoma. J Am Acad Dermatol. 2016; 75(5):967-974.
-
Melanoma Staging the 8th Edition https://acsjournals.onlinelibrary.wiley.com/doi/10.3322/caac.21409