Clinical Skin Lesion Diagnosis Using Representations Inspired by Dermatologist Criteria

The skin is the largest organ in human body. Around 30%-70% of individuals worldwide have skin related health problems, for whom effective and efficient diagnosis is necessary. Recently, computer aided diagnosis (CAD) systems have been successfully applied to the recognition of skin cancers in dermatoscopic images. However, little work has concentrated on the commonly encountered skin diseases in clinical images captured by easily-accessed cameras or mobile phones. Meanwhile, for a CAD system, the representations of skin lesions are required to be understandable for dermatologists so that the predictions are convincing. To address this problem, we present effective representations inspired by the accepted dermatological criteria for diagnosing clinical skin lesions. We demonstrate that the dermatological criteria are highly correlated with measurable visual components. Accordingly, we design six medical representations considering different criteria for the recognition of skin lesions, and construct a diagnosis system for clinical skin disease images. Experimental results show that the proposed medical representations can not only capture the manifestations of skin lesions effectively, and consistently with the dermatological criteria, but also improve the prediction performance with respect to the state-of-the-art methods based on uninterpretable features.

1. Introduction

The skin is directly exposed to the air, which leads to skin disease being one of the most common human illnesses. It pervades all cultures, occurs at all ages, and affects the health of 30% to 70% of individuals, with even higher rates for the at-risk subpopulations [21], [22]. A common technique used by dermatologists for diagnosing skin diseases is the dermoscope which enables observation of the latent structures of skin lesions, i.e., a region suffering from disease, whose effects are otherwise invisible to the naked eye. Dermoscopy is usually used for melanomas and other kinds of skin cancers with pigmentation [13]. However, it is difficult to access a dermoscope in resource-poor regions [20], and moreover it is unnecessary for most of the common skin diseases. Therefore, developing an effective skin disease diagnosis system based on easily accessed clinical images would be beneficial and could provide low-cost, universal access to more people. Figure 1.

Given a patient with a common skin disease (the left box), dermatologists make a diagnosis by observing the appearance of the lesions (the top right box). This paper aims to design a skin disease recognition system (the bottom right box) based on clinical images and dermatological criteria. Firstly, the medical information observed by the doctors during diagnosis is exploited. Then, measurable medical representations for skin lesions are designed for diagnosis.