Deep Learning for Diagnostic Binary Classification of Multiple-Lesion Skin Diseases.

Kenneth Thomsen,Hans Bredsted Lomholt,Ole Winther,Anja Liljedahl Christensen,Lars Iversen

doi:10.3389/fmed.2020.574329

Abstract

Background: Diagnosis of skin diseases is often challenging and computer-aided diagnostic tools are urgently needed to underpin decision making.Objective: To develop a convolutional neural network model to classify clinically relevant selected multiple-lesion skin diseases, this in accordance to the STARD guidelines.Methods: This was an image-based retrospective study using multi-task learning for binary classification. A VGG-16 model was trained on 16,543 non-standardized images. Image data was distributed in training set (80%), validation set (10%), and test set (10%). All images were collected from a clinical database of a Danish population attending one dermatological department. Included was patients categorized with ICD-10 codes related to acne, rosacea, psoriasis, eczema, and cutaneous t-cell lymphoma.Results: Acne was distinguished from rosacea with a sensitivity of 85.42% CI 72.24–93.93% and a specificity of 89.53% CI 83.97–93.68%, cutaneous t-cell lymphoma was distinguished from eczema with a sensitivity of 74.29% CI 67.82–80.05% and a specificity of 84.09% CI 80.83–86.99%, and psoriasis from eczema with a sensitivity of 81.79% CI 78.51–84.76% and a specificity of 73.57% CI 69.76–77.13%. All results were based on the test set.Conclusion: The performance rates reported were equal or superior to those reported for general practitioners with dermatological training, indicating that computer-aided diagnostic models based on convolutional neural network may potentially be employed for diagnosing multiple-lesion skin diseases.

Highlights

Skin diseases rank fourth among non-fatal diseases with respect to global burden [1] and are, estimated to account for 12–20% of general practitioner (GP) consultations [2, 3]
Cutaneous t-cell lymphoma (CTCL) is a rare malignant disease of the skin that is often difficult to distinguish from eczematous disease, even for trained dermatologists [6]
We focused on five multiple-lesion skin diseases and on non-standardized imagery to accommodate the paucity in the scientific literature [8], and—more importantly—to imitate the real-life clinical settings of primary healthcare professionals, where multiplelesion dermatological diseases are often encountered and access to a dermatoscope is limited [20, 21]

Summary

Introduction

Skin diseases rank fourth among non-fatal diseases with respect to global burden [1] and are, estimated to account for 12–20% of general practitioner (GP) consultations [2, 3]. GP diagnostic accuracy in dermatological disease has been estimated to fall in the 48–77% range [5]. For GP’s distinguishing between the two morphologically similar and common papulo-pustular skin diseases of acne and rosacea, and between the two common scaly erythematous diseases of psoriasis and eczema can be a challenge. Cutaneous t-cell lymphoma (CTCL) is a rare malignant disease of the skin that is often difficult to distinguish from eczematous disease, even for trained dermatologists [6]. Diagnosis of skin diseases is often challenging and computer-aided diagnostic tools are urgently needed to underpin decision making

Methods

Results

Conclusion