AElm-XGB: Hybrid Machine Learning Based Automated Classification of Oral Cancer by Effective Segmentation

Abstract

Medical professionals benefit from early cancer disease detection because it shortens the time between diagnosis and treatment. In small towns and rural areas where people are less aware of the risks associated with tobacco and cigarette use, OC may have a high mortality rate due to its severe effects. The technique is insufficient for detecting OC disease in its early stages. Since people only live for five years on average after being diagnosed with OC disease, its detection may be possible at a later stage. By overcoming certain demerits of existing works, the proposed research work presents a novel machine learning (ML) based methodology for accurate diagnosis of oral cancer. At first, the input images are collected and pre-processed using the Quadruple Clipped Enhanced Histogram equalization (QCeHE) for enhancing the image contrast and Adaptive Prewitt and Laplacian of Gaussian filtering (ApLG) for noise removal. After the pre-processing step, hybrid segmentation is done on the basis of Expanded gate attention U-Net model (ExpG-UNet) method. Here the original images will be trained with that of generated ground truth images to obtain the segmented output. The features are extracted based on Integration of discrete wavelet transform, principle component analysis and gray level co-occurrence matrix (DWT-PCA-GLCM) method and extracted features are fused together. The features are provided to the machine learning method called Advanced Hybrid Extreme Learning Machine and Extreme Gradient Boosting (AElm-XGB) to classify the oral cancer effectively. The losses can be optimized by using African Vulture Optimization (AVO).The proposed model can accurately detect the oral cancer effectively. In the evaluation, the proposed model achieved 98.67 % accuracy. Diabetes mellitus, particularly type 2 diabetes, is a growing global health issue, often exacerbated by insulin resistance, metabolic disturbances, and inflammatory responses. Emodin, a natural anthraquinone derivative, has been widely recognized for its pharmacological properties, including its antidiabetic potential. This study investigates the therapeutic efficacy of emodin in ameliorating Type 2 diabetes in Wistar albino rats induced by streptozotocin (STZ). Male rats were divided into healthy, diabetic, and treatment groups, with diabetic groups receiving emodin (40 mg/kg body weight/day, orally) or metformin for 45 days. Diabetes induction was confirmed by elevated blood glucose levels, altered lipid profiles, and reduced insulin sensitivity. Emodin administration significantly reduced fasting blood glucose levels and improved glucose tolerance, comparable to metformin treatment. Biochemical analyses revealed that emodin restored the lipid profile, enhanced antioxidant enzyme activity, and suppressed oxidative stress markers in diabetic rats. The study revealed that emodin exhibited antidiabetic effects by regulating glucose metabolism, enhancing insulin sensitivity, and reducing oxidative stress. Data expressed as mean ± SEM showed significant differences (p<0.05) among control, diabetic control, emodin-treated, and metformin-treated groups. Emodin treatment notably improved glucose metabolism and outperformed metformin in reducing insulin resistance. These findings highlight emodin's potential as a therapeutic agent for Type 2 diabetes by targeting insulin resistance, inflammation, and metabolic dysregulation. Emodin holds promise for managing diabetes and its complications effectively.