Introduction Clinical trials often suffer from significant recruitment barriers, poor adherence, and dropouts, which increase costs and negatively affect trial outcomes. The aim of this study was to examine whether making it virtual and reward-based would enable nationwide recruitment, identify patients with variable disease severity, achieve high adherence, and reduce dropouts. Methods In a siteless, virtual feasibility study, individuals with atopic dermatitis (AD) were recruited online. During the 8-week study, subjects used their smartphones weekly to photograph target AD lesions, and completed patient-oriented eczema measure (POEM) and treatment use questionnaires. In return, subjects were rewarded every week with personalized lifestyle reports based on their DNA. Results Over the course of the 11 day recruitment period, 164 (82% women and 18% men) filled in the form to participate, of which 65 fulfilled the inclusion criteria and signed the informed consent. Ten were excluded as they did not complete the mandatory study task of returning the DNA sample. 55 (91% women, 9% men) subjects returned the DNA sample and were enrolled throughout Denmark, the majority outside the Copenhagen capital region in rural areas with relatively low physician coverage. The mean age was 28.5 (SD ±9.5 years, range 18-52 years). The baseline POEM score was 14.5±5.6 (range 6-28). Based on the POEM, 7 individuals had mild, 28 had moderate, 17 had severe, and 3 had very severe eczema. The retention rate was 96% as 53 out of 55 enrolled completed the study.

Background Digital imaging of dermatological patients is a novel approach to remote assessment and has recently become more relevant since telehealth and remote decentralized clinical trials are gaining ground. Objective We aimed to investigate whether photographs taken by a smartphone are of adequate quality to allow severity assessments to be made and to explore the usefulness of an established atopic dermatitis severity assessment instrument on photograph evaluation. Methods During scheduled visits in a previously published study, the investigating doctor evaluated the severity of atopic dermatitis using the Scoring AD (SCORAD) index and took photographs of the most representative lesions (target lesions) with both a smartphone and a digital single-lens reflex camera (DSLR). The photographs were then assessed by 5 dermatologists using the intensity items of the SCORAD (iSCORAD), which consists of erythema, oedema/papulation, excoriations, lichenification, oozing/crusts, and dryness (scale 0-3, maximum score 18). The mean iSCORAD of the photographs was calculated and compared with in-person assessments using Pearson correlation and Bland-Altman plots. Intraclass correlation coefficients were used for interrater reliability. Results A total of 942 photographs from 95 patients were assessed. The iSCORAD based on smartphone photographs correlated strongly with the evaluations performed in person (iSCORAD: r=0.78, P<.001; objective SCORAD: r=0.81, P<.001; and total SCORAD: r=0.78, P<.001). For iSCORAD specifically, a Bland-Altman plot showed a difference in mean score of 1.31 for in-person and remote iSCORAD. In addition, the interrater agreement between the 5 rating dermatologists was 0.93 (95% CI 0.911-0.939). A total of 170 lesions were photographed, and the difference in mean scores was 1.32, 1.13, and 1.43 between in-person and remote evaluations based on photographs taken by a DSLR camera, a smartphone without flash, and a smartphone with flash, respectively. Conclusions In terms of quality, remote atopic dermatitis severity assessments based on photographs are comparable to in-person assessments, and smartphone photos can be used to assess atopic dermatitis severity to a similar degree as photographs from a DSLR camera. Further, the variation in how the dermatologists in this study rated the iSCORAD based on the photographs was very low.

Background: The use of photographs to diagnose and monitor skin diseases is gaining ground.Objectives: To investigate the validity and reliability of photographic assessments of atopic dermatitis (AD) severity.Methods: AD severity was evaluated in the clinic by two assessors using the Eczema Area and Severity Index (EASI), SCOring Atopic Dermatitis (SCORAD), and Investigator's Global Assessment (IGA). Participants photographed the lesions with their own smartphone and completed a questionnaire about the extent of eczema the same day from home. The photographs were assessed twice with an 8 weeks interval by five dermatologists experienced in photographic evaluations. Intraclass correlation coefficients (ICC) with 95% confidence interval (CI) were applied.Results: Seventy-nine participants were enrolled. The ICC between clinical EASI and photographic EASI was 0.88 (95% CI 0.81-0.93), and 0.86 (0.70-0.93) between clinical SCORAD and photographic SCORAD.Perfect agreement between clinical IGA and photograph IGA was observed for 62%, with the difference between the two never deviating with more than 1 score.The inter-rater ICC for photographic EASI and photographic SCORAD, respectively, was 0.90 (0.85-0.94), and 0.96 (0.91-0.98). The intra-rater agreements between the first and second assessments varied from 0.95 to 0.98 for photographic EASI, and from 0.86 to 0.94 for photographic SCORAD.

No abstract

BACKGROUND Convolutional neural networks (CNNs) are regarded as state-of-the-art artificial intelligence (AI) tools for dermatological diagnosis, and they have been shown to achieve expert-level performance when trained on a representative dataset. CNN explainability is a key factor to adopting such techniques in practice and can be achieved using attention maps of the network. However, evaluation of CNN explainability has been limited to visual assessment and remains qualitative, subjective, and time consuming. OBJECTIVE This study aimed to provide a framework for an objective quantitative assessment of the explainability of CNNs for dermatological diagnosis benchmarks. METHODS We sourced 566 images available under the Creative Commons license from two public datasets—DermNet NZ and SD-260, with reference diagnoses of acne, actinic keratosis, psoriasis, seborrheic dermatitis, viral warts, and vitiligo. Eight dermatologists with teledermatology expertise annotated each clinical image with a diagnosis, as well as diagnosis-supporting characteristics and their localization. A total of 16 supporting visual characteristics were selected, including basic terms such as <i>macule, nodule, papule, patch, plaque, pustule,</i> and <i>scale</i>, and additional terms such as <i>closed comedo, cyst, dermatoglyphic disruption, leukotrichia, open comedo, scar, sun damage, telangiectasia</i>, and <i>thrombosed capillary</i>. The resulting dataset consisted of 525 images with three rater annotations for each. Explainability of two fine-tuned CNN models, ResNet-50 and EfficientNet-B4, was analyzed with respect to the reference explanations provided by the dermatologists. Both models were pretrained on the ImageNet natural image recognition dataset and fine-tuned using 3214 images of the six target skin conditions obtained from an internal clinical dataset. CNN explanations were obtained as activation maps of the models through gradient-weighted class-activation maps. We computed the fuzzy sensitivity and specificity of each characteristic attention map with regard to both the fuzzy gold standard characteristic attention fusion masks and the fuzzy union of all characteristics. RESULTS On average, explainability of EfficientNet-B4 was higher than that of ResNet-50 in terms of sensitivity for 13 of 16 supporting characteristics, with mean values of 0.24 (SD 0.07) and 0.16 (SD 0.05), respectively. However, explainability was lower in terms of specificity, with mean values of 0.82 (SD 0.03) and 0.90 (SD 0.00) for EfficientNet-B4 and ResNet-50, respectively. All measures were within the range of corresponding interrater metrics. CONCLUSIONS We objectively benchmarked the explainability power of dermatological diagnosis models through the use of expert-defined supporting characteristics for diagnosis.