Intellectual assessment of amyotrophic lateral sclerosis using deep resemble forward neural network

• Published in: Neural networks Vol. 178; p. 106478
• Main Authors: Alqahtani, Abdullah, Alsubai, Shtwai, Sha, Mohemmed, Dutta, Ashit Kumar, Zhang, Yu-Dong
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.10.2024
• Subjects: Amyotrophic lateral sclerosis; Classification; Deep learning; Machine learning; Neurotic disorders; Deep learning; Amyotrophic lateral sclerosis; Neurotic disorders; Machine learning; Classification
• ISSN: 0893-6080; 1879-2782; 1879-2782
• DOI: 10.1016/j.neunet.2024.106478

ALS (Amyotrophic Lateral Sclerosis) is a neurodegenerative disorder causing profound physical disability that severely impairs a patient's life expectancy and quality of life. It also leads to muscular atrophy and progressive weakness of muscles due to insufficient nutrition in the body. At present, there are no disease-modifying therapies to cure ALS, and there is a lack of preventive tools. The general clinical assessments are based on symptom reports, neurophysiological tests, neurological examinations, and neuroimaging. But, these techniques possess various limitations of low reliability, lack of standardized protocols, and lack of sensitivity, especially in the early stages of disease. So, effective methods are required to detect the progression of the disease and minimize the suffering of patients. Extensive studies concentrated on investigating the causes of neurological disease, which creates a barrier to precise identification and classification of genes accompanied with ALS disease. Hence, the proposed system implements a deep RSFFNNCNN (Resemble Single Feed Forward Neural Network-Convolutional Neural Network) algorithm to effectively classify the clinical associations of ALS. It involves the addition of custom weights to the kernel initializer and neutralizer ‘k’ parameter to each hidden layer in the network. This is done to increase the stability and learning ability of the classifier. Additionally, the comparison of the proposed approach is performed with SFNN (Single Feed NN) and ML (Machine Learning) based algorithms, namely, NB (Naïve Bayes), XGBoost (Extreme Gradient Boosting) and RF (Random Forest), to estimate the efficacy of the proposed model. The reliability of the proposed algorithm is measured by deploying performance metrics such as precision, recall, F1 score, and accuracy.