AI-powered visual diagnosis of vulvar lichen sclerosus: A pilot study.

BACKGROUND: Vulvar lichen sclerosus (VLS) is a chronic inflammatory skin condition associated with significant impairment of quality of life and potential risk of malignant transformation. However, diagnosis of VLS is often delayed due to its variable clinical presentation and shame-related late consultation. Machine learning (ML)-trained image recognition software could potentially facilitate early diagnosis of VLS. OBJECTIVE: To develop a ML-trained image-based model for the detection of VLS. METHODS: Images of both VLS and non-VLS anogenital skin were collected, anonymized, and selected. In the VLS images, 10 typical skin signs (whitening, hyperkeratosis, purpura/ecchymosis, erosion/ulcers/excoriation, erythema, labial fusion, narrowing of the introitus, labia minora resorption, posterior commissure (fourchette) band formation and atrophic shiny skin) were manually labelled. A deep convolutional neural network was built using the training set as input data and then evaluated using the test set, where the developed algorithm was run three times and the results were then averaged. RESULTS: A total of 684 VLS images and 403 non-VLS images (70% healthy vulva and 30% with other vulvar diseases) were included after the selection process. A deep learning algorithm was developed by training on 775 images (469 VLS and 306 non-VLS) and testing on 312 images (215 VLS and 97 non-VLS). This algorithm performed accurately in discriminating between VLS and non-VLS cases (including healthy individuals and non-VLS dermatoses), with mean values of 0.94, 0.99 and 0.95 for recall, precision and accuracy, respectively. CONCLUSION: This pilot project demonstrated that our image-based deep learning model can effectively discriminate between VLS and non-VLS skin, representing a promising tool for future use by clinicians and possibly patients. However, prospective studies are needed to validate the applicability and accuracy of our model in a real-world setting.

Objective hand eczema severity assessment with automated lesion anatomical stratification.

Hand eczema (HE) is one of the most frequent dermatoses, known to be both relapsing and remitting. Regular and precise evaluation of the disease severity is key for treatment management. Current scoring systems such as the hand eczema severity index (HECSI) suffer from intra- and inter-observer variance. We propose an automated system based on deep learning models (DLM) to quantify HE lesions' surface and determine their anatomical stratification. In this retrospective study, a team of 11 experienced dermatologists annotated eczema lesions in 312 HE pictures, and a medical student created anatomical maps of 215 hands pictures based on 37 anatomical subregions. Each data set was split into training and test pictures and used to train and evaluate two DLMs, one for anatomical mapping, the other for HE lesions segmentation. On the respective test sets, the anatomy DLM achieved average precision and sensitivity of 83% (95% confidence interval [CI] 80-85) and 85% (CI 82-88), while the HE DLM achieved precision and sensitivity of 75% (CI 64-82) and 69% (CI 55-81). The intraclass correlation of the predicted HE surface with dermatologists' estimated surface was 0.94 (CI 0.90-0.96). The proposed method automatically predicts the anatomical stratification of HE lesions' surface and can serve as support to evaluate hand eczema severity, improving reliability, precision and efficiency over manual assessment. Furthermore, the anatomical DLM is not limited to HE and can be applied to any other skin disease occurring on the hands such as lentigo or psoriasis.

Quantification of Efflorescences in Pustular Psoriasis Using Deep Learning.

OBJECTIVES: Pustular psoriasis (PP) is one of the most severe and chronic skin conditions. Its treatment is difficult, and measurements of its severity are highly dependent on clinicians' experience. Pustules and brown spots are the main efflorescences of the disease and directly correlate with its activity. We propose an automated deep learning model (DLM) to quantify lesions in terms of count and surface percentage from patient photographs. METHODS: In this retrospective study, two dermatologists and a student labeled 151 photographs of PP patients for pustules and brown spots. The DLM was trained and validated with 121 photographs, keeping 30 photographs as a test set to assess the DLM performance on unseen data. We also evaluated our DLM on 213 unstandardized, out-of-distribution photographs of various pustular disorders (referred to as the pustular set), which were ranked from 0 (no disease) to 4 (very severe) by one dermatologist for disease severity. The agreement between the DLM predictions and experts' labels was evaluated with the intraclass correlation coefficient (ICC) for the test set and Spearman correlation (SC) coefficient for the pustular set. RESULTS: On the test set, the DLM achieved an ICC of 0.97 (95% confidence interval [CI], 0.97-0.98) for count and 0.93 (95% CI, 0.92-0.94) for surface percentage. On the pustular set, the DLM reached a SC coefficient of 0.66 (95% CI, 0.60-0.74) for count and 0.80 (95% CI, 0.75-0.83) for surface percentage. CONCLUSIONS: The proposed method quantifies efflorescences from PP photographs reliably and automatically, enabling a precise and objective evaluation of disease activity.

Tear Proteomic Predictive Biomarker Model for Ocular Graft Versus Host Disease Classification.

Purpose: Diagnosis of ocular graft-versus-host disease (oGVHD) is hampered by a lack of clinically-validated biomarkers. This study aims to predict disease severity on the basis of tear protein expression in mild oGVHD. Methods: Forty-nine patients with and without chronic oGVHD after AHCT were recruited to a cross-sectional observational study. Patients were stratified using NIH guidelines for oGVHD severity: NIH 0 (none; n = 14), NIH 1 (mild; n = 9), NIH 2 (moderate; n = 16), and NIH 3 (severe; n = 10). The proteomic profile of tears was analyzed using liquid chromatography-tandem mass spectrometry. Random forest and penalized logistic regression were used to generate classification and prediction models to stratify patients according to disease severity. Results: Mass spectrometry detected 785 proteins across all samples. A random forest model used to classify patients by disease grade achieved F1-measure values for correct classification of 0.95 (NIH 0), 0.8 (NIH 1), 0.74 (NIH 2), and 0.83 (NIH 3). A penalized logistic regression model was generated by comparing patients without oGVHD and those with mild oGVHD and applied to identify potential biomarkers present early in disease. A panel of 13 discriminant markers achieved significant diagnostic accuracy in identifying patients with moderate-to-severe disease. Conclusions: Our work demonstrates the utility of tear protein biomarkers in classifying oGVHD severity and adds further evidence indicating ocular surface inflammation as a main driver of oGVHD clinical phenotype. Translational Relevance: Expression levels of a 13-marker tear protein panel in AHCT patients with mild oGVHD may predict development of more severe oGVHD clinical phenotypes.