Artificial intelligence for the automated single‐shot assessment of psoriasis severity

Abstract

Background PASI score is globally used to assess disease activity of psoriasis. However, it is relatively complicated and time-consuming, and the score will vary due to the inconsistent subjectivity between dermatologists. Therefore, an AI system capable of assessing psoriasis severity will be useful.Objectives To propose a simplified PASI system (Single-Shot PASI) and associated AI models capable of assessing psoriasis severity.Methods Overall, 705 psoriasis images of the trunk's front and back were used in our research. Considering the relatively small number of images, we used data augmentation techniques to expand the data. A psoriasis expert's scores were used as teacher data. Various convolutional neural network models and hyperparameters were adjusted using a fivefold cross-validation. From these adjustments, we discovered that fine-tuning Imagenet2012-pretrained InceptionV3 whose last linear layer was replaced by a two-layer perceptron (30 hidden units and five output units) exhibited the best performance. ResultsTo validate our deep learning system, 10 images were selected as test sets and were excluded from the training sets. The AI assessment of Single-Shot PASI was almost consistent with the clinical severity. We examined whether AI assistance would affect human scoring. In this study, 13 dermatologists and 9 medical students were invited as evaluators. Mean absolute differences from AI scores and standard deviation among evaluators reduced with AI assistance.In addition, the evaluator's scores got close to the teacher's score with AI's assistance.Conclusions We proposed a Single-Shot PASI system and developed an associated AI system capable of assessing psoriasis severity simply by uploading a single clinical image. An easy-to-use scoring system and our freely available AI software would help dermatologists and patients with psoriasis.

Artificial intelligence in psoriasis: Where we are and where we are going

“…Dermatologists currently grade psoriasis severity based on the BSA and PASI 54 . However, relying on subjective evaluations of dermatologists may lead to deviations in disease diagnosis 55 . Therefore, there is a need for an automated system to evaluate psoriasis lesion severity and affected area in the diagnosis and treatment of psoriasis.…”

“…For erythema, scaliness, and induration scoring, CNNs performed similarly to dermatologists, while for BSA scoring, CNNs outperformed dermatologists on image‐based PASI scoring 63 . Building on this, a simplified PASI system (Single‐Shot PASI) was developed, which can assess psoriasis severity simply by uploading a clinical image and can closely approximate dermatologist evaluation 55 . Recently, a portable device has been clinically employed for automated PASI measurements after total body imaging and digital image analysis.…”

“…54 However, relying on subjective evaluations of dermatologists may lead to deviations in disease diagnosis. 55 Therefore, there is a need for an automated system to evaluate psoriasis lesion severity and affected area in the diagnosis and treatment of psoriasis. This can be achieved through segmentation of psoriasis lesions in skin images from psoriasis patients, which enables the automation of lesion severity and affected area evaluation.…”

“…63 Building on this, a simplified PASI system (Single-Shot PASI) was developed, which can assess psoriasis severity simply by uploading a clinical image and can closely approximate dermatologist evaluation. 55 Recently, a portable device has been clinically employed for automated PASI measurements after total body imaging and digital image analysis. This device has shown high reproducibility and agreement with results obtained by PASI-trained dermatologists in pilot clinical validation.…”

Automated dermatoscopic pattern discovery by clustering neural network output for human–computer interaction

“…In the field of dermatology, CNNs were described to have at least an equal accuracy as dermatologists in experimental settings for classifying clinical and dermatoscopic images, and shown to improve physicians' diagnostic accuracy when applied in diverse interactive settings [8], [9]. Such algorithms can not only classify images but also label anatomic areas [10], rate psoriasis [11], or retrieve similar images to a case, by implicitly analyzing patterns and pattern combinations after training to categorize images into distinct classes [12]. Therefore, we hypothesize that convolutional neural networks could be helpful in the extraction of diagnostically relevant patterns in medical image collections.…”