In this chapter, the different dermatological diseases are differentiated using the concept of probabilistic neural networks (PNN). There are different colour transformations for various dermatological diseases. This chapter mainly focuses on the differentiation of psoriasis and dermatitis with the normal one. The colour, shape, and textural features of the patches are studied and analysed for recognising the various dermatological diseases. The colour, texture, mean, median, entropy, standard deviations are considered for feature analysis. The pre-screening system uses the PNN algorithm after the feature extraction. The experimental results define the accuracy and effectiveness of the proposed algorithm, and the system scores sensitivity of 0.91, specificity of 0.94, with an accuracy rate of 96.25%.
TopIntroduction
Dermatological diseases have major impacts and concerns on people’s lives and are one of the most prevalent diseases which people are taken into consideration. According to the survey mentioned in Karimkhani et al. (2017) and Hay et al. (2010), almost one-third of the worldwide population suffer from skin diseases. Sometimes, the early diagnosis of dermatological disease is difficult, and it may enhance the severity if it is left undetected. Skin disease can be caused due to various reasons which include heredity, exposure to chemicals, lack of cleanliness, usage of cigarettes, drugs, alcohol, etc., and even the climatic change is also one of the factors. Though it does not have much relation with mortality rates in connection with dermatological abnormalities, these diseases have greater impacts on the quality of life. The skin covers the entire body and it acts as a protective barrier layer against the damage to internal tissues. It is having a total of 20 square feet of area with a weight of around 8 pounds. The detailed illustration of the skin anatomy is depicted in figure 1.
The skin has mainly three layers: Epidermis, Dermis and Hypodermis. The skin colouring pigment called melanocytes are located in the epidermis layer, the outermost part of our skin, and this layer creates the texture. Most of the symptoms are shown in this layer. The dermis layer contains tough connective tissues, hair follicles, sweat glands. Much deeper, there is another layer is called the Hypodermis layer which contains fat and other connective tissues.
People have a weak knowledge about the diseases, that they consult a dermatologist several weeks or months later after the skin irritation starts, causing the enhancement in the disease severity. In some cases, the doctors find it difficult to diagnose the disease symptoms at its early stage and it requires expensive tests in order to diagnose the disease and its severity level. Though lasers and photonics technology are available, the expensiveness of such laboratory tests is still a limitation. Here, the authors are proposing a neural network-based image processing algorithm, capable of correctly diagnosing the disease and its stage. This system can be effectively used by the dermatologists for their reference. Another peculiarity of the proposed system is that, it helps the patient to diagnose whether the abnormalities on the skin are the symptoms for normal flakes or rashes, by imputing the captured image taken from a smartphone into the Probabilistic Neural Network algorithm. Some of the skin infections include Rashes, Dermatitis, Psoriasis, Eczema, Acne, etc., even some of the abnormalities have the same visual characteristics.