Акценти от 3-ти ден

Artificial intelligence (AI) and deep learning in dermatology, New in melasma, The therapeutic care of Merkel cell carcinoma, CAR-T cell therapy for cancer, Sun protection


Dr. Nicole Jouan

Dermatologist, Brest, France

Predicting the future: Artificial intelligence (AI) and deep learning in dermatology

Dr. Allan C. Halpern, Dermatologist, United States
Dr. Roger Ho, Dermatologist, United States
Dr. Adewole Shomari Adamson, Dermatologist, United States
Dr. Jeffrey A. V. Benabio, Dermatologist, United States

Between traditional programming and artificial intelligence, there lies a revolution: the emergence, in 2012, of deep learning, made possible by the increase in the computational power of computers, the invention of convolutional neural networks, and the surge in big data.

This revolutionary technology enabled Google’s AI program Alphago to beat Lee Sedol, the world's 3rd best Go player, in 2016, and allowed AlphaZero to beat the Stockfish 8 chess programme after only four hours of self-training in 2017. Thus, computers today do more than just practise and perform: they also learn, which is referred to as machine learning. This involves programming a computer to make it capable of autonomously learning from databases.

One of the most significant forms of this AI is image recognition. ImageNet is a database that contains more than 15 million images divided into 22,000 categories. It enables algorithms to be trained to recognise all sorts of objects, animals, faces, etc. Facial recognition is already used by Amazon to avoid check-out lines. It is also used by China to “track” its citizens. In dermatology, what first comes to mind is the diagnosis of tumours. Dr Allan Halpern from NYC presented the International Skin Imaging Collaboration (ISIC), justifying its existence: image analysis algorithms are trained using ImageNet databases, i.e. with photos taken in variable conditions. To improve the relevance of the algorithms for the diagnosis of tumours (or any other skin disease), it is thus necessary to enhance the databases with standardised photos (type of camera, lighting, zoom, etc.). He adamantly calls on the scientific community to take up this issue, keeping in mind that our computers contain millions of untapped photos with a corresponding diagnosis. Every year, the ISIC organises the ISIC challenge, which is a competition intended for computer programmers around the world. Its aim is to improve melanoma diagnostic algorithms. The ISIC's archives contain the largest public collection of dermoscopic images of skin lesions with a validated diagnosis (more than 13,000). The participants train their algorithms with this image bank. The algorithms are compared with the responses of expert dermoscopy dermatologists. Remember the first publication of this type by a Stanford team in Nature in 2017, which made big waves in the American press: “Stanford's Artificial Intelligence Is Nearly as Good as Your Dermatologist” (Fortune, January 2017).

These developments are raising many questions addressed by Dr Ho, namely: who should be entrusted with these diagnostic systems?

Doctors? Certainly, to assist with difficult diagnoses.

Medical assistants for “triage”?

Patients themselves?

The risk is an increase in “overdiagnoses” for low-risk lesions and the blockage of already-fragile healthcare systems. We are not yet at the stage of booths acquiring whole-body images to automatically target high-risk lesions and only send dermatologists those requiring removal.

Dr Adamson emphasised the lack of ethnic diversity in photo databases intended for image recognition.

Remember that in 2015, the face and object recognition system of the Google Photos services misidentified a young black woman as a gorilla: if nothing is done, it seems likely that AI will reinforce social inequalities and leave aside a whole portion of the population. The solution: increase racial diversity in databases or create specific algorithms for dark skin.

After addressing the issue of responsibility (can an algorithm be sued?), the development of algorithms for the diagnosis of diabetic retinopathy, colon polyps, lymph node metastases in breast cancer, etc., Dr Benabio ended this fascinating session with a stirring homage to medicine and doctors. He affirmed that just like Spotify will never replace an evening at La Scala, AI cannot replace the special contact between doctors and patients. We will use AI in our profession, but we remain responsible human beings with freedom of choice.

Dr. Rémi Maghia

Dermatologist, Brest, France

What's new in melasma?

Dr. Seemal R. Desai, Dermatologist, United States

The usual therapeutic options are: topical retinoids and combination therapy, azelaic acid, hydroquinone, chemical peels, cosmeceuticals, lasers, dermabrasion.

Triple combination therapy

A randomised, multi-centre, placebo-controlled Asian study (Chan, BJD 2008) compared 260 patients, including 129 treated with triple combination therapy and 131 treated with hydroquinone alone: triple combination therapy was superior to monotherapy in terms of efficacy, but there were more side effects.

Tranexamic acid (TXA)

TXA is a fibrinolytic agent that inhibits the plasminogen activator. Less arachidonic acid results in fewer prostaglandins leading to reduced tyrosinase activity, ultimately reducing pigmentation.

Contraindications: history of thrombosis, anticoagulant treatments, pregnancy, breastfeeding, smoking, and renal, cardiac and pulmonary diseases.

The most effective dose for melasma seems to be 250 to 500 mg per day.

There are topical, oral and intradermal forms.

Tranexamic acid (TXA) was recently introduced as topical therapy to reduce melasma pigmentation, hyperpigmentation and post-inflammatory hyperpigmentation.

TXA's mechanism of action is as follows: it regulates pigmentation by inhibiting the release of inflammatory mediators, specifically prostaglandins and arachidonic acid, which are involved in melanogenesis.

In vitro, TXA inhibits PGE-2 melanin production induced by human epidermal melanocytes.

In trials, TXA led to significant clinical improvement in discoloration biomarkers. The author cited his article to be published in May 2019 in JDD and the Taraz et al. article in Dermatologic Therapy in 2013.

Azelaic acid (AA)

Inhibits tyrosinase and mitochondrial respiration enzymes.

20% was used twice a day in 132 patients with facial melasma: 73% had improvement after six months of treatment. Minimal side effects: erythema, pruritus, moderate burning sensation.


Cysteamine is physiologically synthesised by all mammalian cells and is a natural antioxidant.

It inhibits tyrosinase and peroxidase, blocks dopaquinone, increases intra-cellular glutathione and lightens black melanin in the stratum corneum via an antioxidant effect.

The author reported the first randomised, placebo-controlled trial in 2015 and the Fahrsi 2017 JTD trial, confirming significant efficacy in reducing melanin within melasma lesions.


Glycolic acid 30% to 50%: five or six peels every two or three weeks.

Salicylic acid 15% to 30%: five or six peels every two or three weeks.

Trichloroacetic acid (TCA) 10% to 15%: we can start with a peel with glycolic acid, then continue with a TCA peel, although we directly use 10% to 15% for phototypes IV to VI.

Mandelic acid: this is a large molecule that penetrates the epidermis very slowly and evenly. It is excellent for sensitive skin.

Dr. Lise Boussemart

Dermatologist, Rennes, France

Dear all,

Тhe third day of the AAD Meeting is coming to an end, and it seems that my husband, who stayed at home alone with our three children, has survived this far. I thus have peace of mind as I start to write today's summary. The 3rd day of the AAD Meeting, which was already in full swing at 7:30 am, despite it being Sunday.

As is often the case with the plenary sessions of these American meetings, the speakers chosen were brilliant and sometimes moving and received awards for their scientific achievements.

The first speaker was Dr Lushniak, who reviewed the highlights of his career, all while providing insight into our noble medical profession, which is truly fantastic and multifaceted. We treat patients, we do research, we educate, we serve the community, and we must not forget that we have the ability to make the world a better place.

Dr Olbricht, the outgoing AAD President, with 40 years of dermatological experience under her belt, seemed concerned about the numerous transformations affecting our speciality in the age of digital technology, the Internet and artificial intelligence. She cited Paul Valéry: “The trouble with our times is that the future is not what it used to be”. She had not imagined should would ever witness such a revolution as that taking place today. The system’s transformation is accelerating, with remote consultations, connected devices and patients who refuse to wait. She encourages all dermatologists to jump on the bandwagon, actively integrate new technologies, and “embrace” the future as it is unfolding, or else our speciality and artificial intelligence diagnostic techniques will evolve like taxis and Uber!

Dr Nghiem, from Mass General Hospital, presented breakthroughs in the therapeutic care of Merkel cell carcinoma, both in the primary tumour stage and in the metastatic stage. These new developments are moving in the right direction for patients, with treatments that are less aggressive and more effective than before. Merkel cell carcinoma remains rare but its incidence continues to rise due to an ageing population and the resulting immunosenescence. It is caused either by Merkel cell polyomavirus (80% of cases in the US, fewer in Australia) or by UV-induced mutations (20% of cases in the US, more in Australia). In all cases, the cancer-causing virus as well as UV rays generate “non-self” antigens that can be recognised by an effective immune system. But with Merkel cell carcinoma, the intra-tumoral T-cell infiltrate often shows exhaustion markers such as PD-1. Since it is so rare, community dermatologists see relatively few cases of it, whereas its mortality rate is five times that of melanoma. Dr Nghiem probably has the largest cohort of Merkel cell carcinoma patients in the world and thus a unique perspective of this disease. He recommends discussing each case in a multidisciplinary consultation meeting. Experience in his clinic has led to abandoning the recommendation of wide margins (2cm) for the excision of a Merkel cell carcinoma with poor prognosis (with sentinel lymph node technique), to avoid creating a flap or graft that would further delay adjuvant radiotherapy. For him, a 1cm margin is sufficient and direct closure should be favoured, with a view to rapid adjuvant radiotherapy. For patients testing positive for the polyomavirus, changes in antibody levels over time have high prognostic value. Indeed, if there is a decline in levels three months following surgery, there will be no recurrence for 98% of patients, which is why Dr Nghiem does not prescribe a follow-up scan in this case. If there is an increase in levels after three months, there will be recurrence for 99% of patients, and close monitoring via imaging is warranted. In the metastatic stage, immunotherapy with PD-1/PD-L1 inhibitors is now the first line of treatment in the US (NCCN recommendations, 2018) with much longer-lasting efficacy and much better tolerance than with older chemotherapies.

It was then time for Dr Crystal Mackall from Stanford to be honoured for her work in CAR-T cell therapy for cancer. CAR-T cells are the patient's own T cells that have been genetically modified in vitro to recognise a target marker in cancer cells to be eliminated. Why bother transforming a patient's T cells when anti-PD-1 immunotherapy (2018 Nobel Prize in Medicine) is now capable of healing cancer simply by injecting antibodies? Well, anti-PD-1 immunotherapy is indeed effective for immunogenic cancers (with a high mutation load or oncogenic virus), but it is not very useful for cancers with few mutations, which is generally the case of childhood leukaemia. With CAR-T cell therapy, the patient's T cells are modified to express a defined chimeric antigen receptor, e.g. CD19, to recognise leukaemia cells. This effective treatment (Lee et al., Lancet 2014) is now the first cell therapy approved by the FDA for resistant or recurring B-cell acute lymphoblastic leukaemia in children. What about solid tumours? In 2018, Dr Mackall demonstrated, in mice, that CAR-T cells recognising the GD2 antigen were effective for a sub-type of paediatric diffuse glioma (Mount et al., Nature Medicine 2018). Cell therapy, with the genetic modification of the patient’s own lymphocytes, was long considered as unrealistic in terms of cost (several hundred thousand euros per treatment) and local technical feasibility/reproducibility. However, the automation/miniaturisation of cell incubators has started to significantly lower costs, and therefore companies such as GSK have concretely invested in this promising solution. Why not imagine CAR-T cells for melanomas with low mutation rates such as plantar, mucosal or choroidal melanomas?

Dr. Steven Wang, Dermatologist, United States

The afternoon session devoted to sun protection partly overlapped with that described yesterday by my co-reporter Dr Rémi Maghia (02/03/19). This was undoubtedly due to current events: the FDA officially proposed a new regulation for solar filters on 26/02/19, i.e. less than a week ago. For a 90-day period, all Americans had the ability to comment on this proposal, which will cause many questions to be asked in dermatological practices. In this proposal, solar filters are classified into three groups: “good” (zinc oxide or ZnO and titanium oxide or TiO2), “bad” (already removed from all American creams: para-aminobenzoic acid or PBA and trolamine salicylate) and “still undetermined” (including oxybenzone, a UVB+UVA filter that is found in 70% of American sun creams and to which 97% of Americans are exposed). This filter has been used since 1978. Endocrine toxicity has been reported in rats at high doses, but to reach such doses, it would be necessary to apply 2mg/cm2 of sun cream every day to the entire body for 34 years. That said, oxybenzone is still capable of causing contact dermatitis and photoallergies. A clinical trial, funded by the FDA (NCT03582215), recently measured the systemic absorption of four solar filters following application. The results, currently submitted for publication, show that of the four tested filters (avobenzone, oxybenzone, ecamsule and octocrylene), oxybenzone had the highest blood concentrations following application. Since this filter passes into the bloodstream and is not removed by standard wastewater treatments, additional studies will be necessary to affirm its real biological safety. Dr Steven Wang nonetheless reiterated that sun creams are an effective method for the primary prevention of sunburn and skin cancer, and in real life, SPF100 creams provide significantly better protection than SPF50 creams (Williams et al., JAAD 2018).