AI-Powered Deep Learning Tool Accurately Assesses Nail Psoriasis Severity: Research Shows
Germany: A team of researchers from Germany developed a deep learning model using a convolutional neural network (CNN) to accurately assess nail psoriasis severity. The model was trained on 4,400 images and demonstrated strong performance, achieving an AUROC of 86% during training and 80% in validation, closely aligning with expert evaluations.
"By incorporating the modified Nail Psoriasis Severity Index, this tool offers a reliable and automated method for assessing nail psoriasis, with significant potential for clinical application," the researchers wrote in Frontiers in Medicine.
Nail psoriasis, a common manifestation of psoriatic disease, often requires precise assessment for effective management. Traditional scoring methods rely on manual evaluation, which can be time-consuming and subject to variability. To address these challenges, Stephan Kemenes, Department of Dermatology, Friedrich-Alexander-University Erlangen-Nürnberg (FAU) and Universitätsklinikum Erlangen, Erlangen, Germany, and colleagues aimed to enhance and validate a CNN-based model for automated nail psoriasis severity scoring using the modified Nail Psoriasis Severity Index (mNAPSI), ensuring accurate assessment across all severity levels without reliance on standardized conditions.
For this purpose, the researchers included patients with psoriasis (PsO), psoriatic arthritis (PsA), and non-psoriatic controls, including healthy individuals and those with rheumatoid arthritis, for model training. Validation was conducted on an independent cohort of psoriatic patients. Nail photographs were pre-processed and segmented, with mNAPSI scores annotated by five expert readers.
A CNN based on the BEiT architecture, pre-trained on ImageNet-22k, was fine-tuned for mNAPSI classification. Model performance was evaluated against human annotations using AUROC and other metrics, while a reader study assessed inter-rater variability.
The key findings were as follows:
- The training dataset included 460 patients who provided 4,400 nail photographs.
- An independent validation dataset consisted of 118 patients contributing 929 nail photographs.
- The CNN showed strong classification performance, achieving a mean AUROC of 86% ± 7% across mNAPSI classes in the training dataset.
- Performance remained stable in the validation dataset, with a mean AUROC of 80% ± 9%, despite variations in imaging conditions.
- Compared to human annotation, the CNN achieved a Pearson correlation of 0.94 at the patient level, maintaining consistency in the validation dataset.
The researchers stated, "We developed and validated a CNN that reliably automates the scoring of nail psoriasis severity using mNAPSI without requiring image standardization."
They concluded, "This method offers potential clinical benefits by enabling a standardized and time-efficient assessment of nail involvement in psoriatic disease and may also serve as a self-reporting tool."
Reference:
Kemenes, S., Chang, L., Schlereth, M., Minopoulou, I., Fenzl, P., Corte, G., Yalcin Mutlu, M., Höner, M. W., Sagonas, I., Simon, D., Kleyer, A., Folle, L., Sticherling, M., Schett, G. A., Maier, A. K., & Fagni, F. Advancement and Independent Validation of a Deep Learning-based tool for Automated Scoring of Nail Psoriasis Severity Using the Modified Nail Psoriasis Severity Index. Frontiers in Medicine, 12, 1574413. https://doi.org/10.3389/fmed.2025.1574413
