Consistency of convolutional neural networks in dermoscopic melanoma recognition: A prospective real-world study about the pitfalls of augmented intelligence.

BACKGROUND: Deep-learning convolutional neural networks (CNNs) have outperformed even experienced dermatologists in dermoscopic melanoma detection under controlled conditions. It remains unexplored how real-world dermoscopic image transformations affect CNN robustness. OBJECTIVES: To investigate the consistency of melanoma risk assessment by two commercially available CNNs to help formulate recommendations for current clinical use. METHODS: A comparative cohort study was conducted from January to July 2022 at the Department of Dermatology, University Hospital Basel. Five dermoscopic images of 116 different lesions on the torso of 66 patients were captured consecutively by the same operator without deliberate rotation. Classification was performed by two CNNs (CNN-1/CNN-2). Lesions were divided into four subgroups based on their initial risk scoring and clinical dignity assessment. Reliability was assessed by variation and intraclass correlation coefficients. Excisions were performed for melanoma suspicion or two consecutively elevated CNN risk scores, and benign lesions were confirmed by expert consensus (n = 3). RESULTS: 117 repeated image series of 116 melanocytic lesions (2 melanomas, 16 dysplastic naevi, 29 naevi, 1 solar lentigo, 1 suspicious and 67 benign) were classified. CNN-1 demonstrated superior measurement repeatability for clinically benign lesions with an initial malignant risk score (mean variation coefficient (mvc): CNN-1: 49.5(±34.3)%; CNN-2: 71.4(±22.5)%; p = 0.03), while CNN-2 outperformed for clinically benign lesions with benign scoring (mvc: CNN-1: 49.7(±22.7)%; CNN-2: 23.8(±29.3)%; p = 0.002). Both systems exhibited lowest score consistency for lesions with an initial malignant risk score and benign assessment. In this context, averaging three initial risk scores achieved highest sensitivity of dignity assessment (CNN-1: 94%; CNN-2: 89%). Intraclass correlation coefficients indicated 'moderate'-to-'good' reliability for both systems (CNN-1: 0.80, 95% CI:0.71-0.87, p < 0.001; CNN-2: 0.67, 95% CI:0.55-0.77, p < 0.001). CONCLUSIONS: Potential user-induced image changes can significantly influence CNN classification. For clinical application, we recommend using the average of three initial risk scores. Furthermore, we advocate for CNN robustness optimization by cross-validation with repeated image sets. TRIAL REGISTRATION: ClinicalTrials.gov (NCT04605822).

Preferences of German and Swiss melanoma patients for toxicities versus melanoma recurrence during adjuvant treatment (GERMELATOX-A-trial).

PURPOSE: Adjuvant treatment with immune checkpoint inhibitors like PD1-antibodies (ICI) ± CTLA4-antibodies (cICI) or targeted therapy with BRAF/MEK inhibitors (TT) in high-risk melanoma patients demonstrate a significant improvement in disease-free survival (DFS). Due to specific side effects, the choice of treatment is very often driven by the risk for toxicity. This study addressed for the first time in a multicenter setting the attitudes and preferences of melanoma patients for adjuvant treatment with (c)ICI and TT. METHODS: In this study ("GERMELATOX-A"), 136 low-risk melanoma patients from 11 skin cancer centers were asked to rate side effect scenarios typical for each (c)ICI and TT with mild-to-moderate or severe toxicity and melanoma recurrence leading to cancer death. We asked patients about the reduction in melanoma relapse and the survival increase at 5 years they would require to tolerate defined side-effects. RESULTS: By VAS, patients on average valued melanoma relapse worse than all scenarios of side-effects during treatment with (c)ICI or TT. In case of severe side effects, patients required a 15% higher rate of DFS at 5 years for (c)ICI (80%) compared to TT (65%). For survival, patients required an increase of 5-10% for melanoma survival during (c)ICI (85%/80%) compared to TT (75%). CONCLUSION: Our study demonstrated a pronounced variation of patient preferences for toxicity and outcomes and a clear preference for TT. As adjuvant melanoma treatment with (c)ICI and TT will be increasingly implemented in earlier stages, precise knowledge of the patient perspective can be helpful for decision making.

Association of sex and systemic therapy treatment outcomes in psoriasis: a two-country, multicentre, prospective, noninterventional registry study.

BACKGROUND: Few systematic data on sex-related treatment responses exist for psoriasis. OBJECTIVES: To evaluate sex differences with respect to systemic antipsoriatic treatment. METHODS: Data from patients with moderate-to-severe psoriasis in the PsoBest or Swiss Dermatology Network of Targeted Therapies (SDNTT) registries were analysed. Treatment response was defined as achieving a ≥ 75% reduction in Psoriasis Area and Severity Index (PASI 75) or PASI ≤ 3 at treatment months 3, 6 and 12, supplemented by patient-reported outcomes [i.e. Dermatology Life Quality Index (DLQI) ≤ 1 and delta DLQI ≥ 4]. RESULTS: In total, 5346 patients registered between 2007 and 2016 were included (PsoBest, n = 4896; SDNTT, n = 450). The majority received nonbiological treatment (67·3% male, 69·8% female). Women showed slightly higher PASI response rates after 3 (54·8% vs. 47·2%; P ≤ 0·001), 6 (70·8% vs. 63·8%; P ≤ 0·001) and 12 months (72·3% vs. 66·1%; P ≤ 0·004). A significantly higher proportion of women achieved a reduction in DLQI ≥ 4 [month 3: 61·4% vs 54·8% (P ≤ 0·001); month 6: 69·6% vs. 62·4% (P ≤ 0·001); month 12: 70·7% vs. 64·4% (P ≤ 0·002)]. Regarding PASI ≤ 3, women on biologics showed a significantly superior treatment response compared with men at 3 (57·8% vs. 48·5%; P ≤ 0·004) and 6 months (69·2% vs. 60·9%; P ≤ 0·018). Women in the nonbiological treatment group had a significantly better treatment response (PASI response, PASI 75 and PASI ≤ 3) over 12 months compared with men. CONCLUSIONS: We provide evidence that women experience better treatment outcomes with systemic antipsoriatic therapy than men.

[Artificial intelligence to support telemedicine in Africa]. / Künstliche Intelligenz zur Unterstützung der Telemedizin am Beispiel Afrikas.

Telemedicine has been used in the daily routine of dermatologists for decades. The potential advantages are especially obvious in African countries having limited medical care, long geographical distances, and a meanwhile relatively well-developed telecommunication sector. National and international working groups support the establishment of teledermatological projects and in recent years have increasingly been using artificial intelligence (AI)-based technologies to support the local physicians. Ethnic variations represent a challenge in the development of automated algorithms. To further improve the accuracy of the systems and to be able to globalize, it is important to increase the amount of available clinical data. This can only be achieved with the active participation of local health care providers as well as the dermatological community and must always be in the interest of the individual patient.

[Role of artificial intelligence in assessing the extent and progression of dermatoses]. / Stellenwert der künstlichen Intelligenz zur Ausbreitungsdiagnostik und Verlaufsbeurteilung von Dermatosen.

BACKGROUND: In recent years, many medical specialties with a visual focus have been revolutionized by image analysis algorithms using artificial intelligence (AI). As dermatology belongs to this field, it has the potential to play a pioneering role in the use of AI. OBJECTIVE: The current use of AI for the diagnosis and follow-up of dermatoses is reviewed and the future potential of these technologies is discussed. MATERIALS AND METHODS: This article is based on a selective review of the literature using Embase and MEDLINE and the keywords "psoriasis", "eczema", "dermatoses" and "acne" combined with "artificial intelligence", "machine learning", "deep learning", "neural network", "computer-guided", "supervised machine learning" or "unsupervised machine learning" were searched. RESULTS: In comparison to examiner-dependent intra- and interindividually fluctuating scores for the assessment of inflammatory dermatoses (e.g. the Psoriasis Areas Severity Index [PASI] and body surface area [BSA]), AI-based algorithms can potentially offer reproducible, standardized evaluations of these scores. Whereas promising algorithms have already been developed for the diagnosis of psoriasis, there is currently only scarce work on the use of AI in the context of eczema. CONCLUSIONS: The latest developments in this field show the enormous potential of AI-based diagnostics and follow-up of dermatological clinical pictures by means of an autonomous computer-based image analysis. These noninvasive, optical examination methods provide valuable additional information, but dermatological interaction remains indispensable in daily clinical practice.