A hybrid machine learning model for skin disease classification using discrete wavelet transform and gray level co-occurrence matrix (GLCM)

• Published in: Multimedia tools and applications Vol. 84; no. 14; pp. 12835 - 12853
• Main Authors: Verma, Sarvachan, Kumar, Manoj
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.04.2025; Springer Nature B.V
• Subjects: Algorithms; Classification; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Decision trees; Discrete Wavelet Transform; Hair removal; Homogeneity; Image filters; Machine learning; Medical imaging; Multimedia Information Systems; Skin diseases; Special Purpose and Application-Based Systems; Supervised learning; Support vector machines; Track 2: Medical Applications of Multimedia; Wavelet transforms; Hybrid classifier; Feature extraction; Gaussian filtering; Machine learning
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-024-19449-5

Skin conditions are widespread health issues in the world. The dangers of infections are hidden and may lead to mental depression and physical health problems. In severe circumstances, it might even result in skin cancer. Therefore, one of the most difficult challenges in medical image handling is identifying skin diseases from clinical images. However, diagnosing skin disease manually by medical professionals is time-consuming and arbitrary. Hence, automatic skin disease prediction is needed by both patients and dermatologists, which makes the treatment strategy fast. This paper core purpose is to predict benign or malignant skin disease. In this paper, we present a technique for hair removal based on morphological filtering, including the Black-Hat transformation to remove small objects, like hair. Then, to eliminate noise from the image, we employ Gaussian filtering. Then discrete wavelet transform ( DWT) algorithm is used for decomposes the image into four sub-band (LL LH, HL, HH); after that, for each channel of the RGB image, A Gray Level Co-occurrence Matrix (GLCM) is calculated to extract the 48 features of the individual image given contrast, correlation, energy, and homogeneity on low frequency (approximate coefficient-LL) and high frequency (horizontal-LH, diagonal-HH, vertical-HL) bands of the segmented image. This Paper focuses on building a hybrid ensemble machine learning model (HEML) (15 Weak learners of Decision Tree (DT), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM) for automatic skin disease classification and comparing the accuracy with other supervised machine learning such as Decision Tree (DT), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM) classifiers. The Hybrid ensemble machine learning model (HEML) model has the highest accuracy of 95.50% compared to the rest of the classification techniques. In addition, we have complemented our work with state-of-the-art methodologies.