Classification and diagnosis of embryonal tumor from microarrays using non-negative matrix factorization
Authors: Quraishi, F.F. and Wani, M.A.
Journal: International Journal of Information Technology Singapore
eISSN: 2511-2112
ISSN: 2511-2104
DOI: 10.1007/s41870-025-02868-4
Abstract:Medulloblastoma, a form of embryonal tumor in the Central Nervous System, is categorized as the most common brain tumor and is mostly diagnosed in children. Biologically, a lot is unknown about medulloblastoma viz., their complete pathogenesis is not discovered. Diagnosing medulloblastoma embryonal tumors using gene expression data from Microarrays is extensively researched but it is very challenging. The factors that are responsible for its challenging nature include class-imbalance nature and high dimensionality associated with Microarrays. Researchers have presented various algorithms that lower the dimensions of the microarray dataset. However, these methods result in over-fitting and poor generalization. A new method is proposed here that is based on the technique of Non-Negative Matrix Factorization (NMF). NMF is good for the reduction of dimensions in microarrays due to the non-negative nature of the dataset. The proposed method improves the accuracy of determining the outcome of patients (survival/failure) in response to therapy. The performance of the algorithm is assessed by using metrics like accuracy, precision, recall, and F-measure. The algorithm is compared with eight well-known algorithms which include: Least Absolute Shrinkage And Selection Operator, Random Forest, Decision Trees, Sine–cosine algorithm, Binary Particle Swarm Optimization, Binary Genetic Algorithm, Differential Equation and PSO-Ensemble using results obtained by employing benchmark datasets. The results indicate that the accuracy and efficiency of the proposed algorithm are better than well-known algorithms.
Source: Scopus