Unsupervised feature dimension reduction for classification of MR spectra

Abstract

We present an unsupervised feature dimension reduction method for the classification of magnetic resonance spectra. The technique preserves spectral information, important for disease profiling. We propose to use this technique as a preprocessing step for computationally demanding wrapper-based feature subset selection. We show that the classification accuracy on an independent test set can be sustained while achieving considerable feature reduction. Our method is applicable to other classification techniques, such as neural networks, support vector machines, etc.

Cost-sensitive regression learning on small dataset through intra-cluster product favoured feature selection

“…Micallef et al (2017) presented a novel approach using an interactive elicitation of knowledge on the feature relevance to guide the selection of features and thus improve the prediction accuracy for small datasets. Another group of studies classifies features into clusters and replaces the feature set with the corresponding feature centroid of the cluster (Baumgartner et al, 2004;Onan, 2017).…”

“…Learning on small dataset Increasement of sample size Niyogi et al (1998), Li et al (2003), Li and Wen (2014), Chang et al (2014). Dimensionality reduction Baumgartner et al (2004), Kursa and Rudnicki (2010), Chandrashekar and Sahin (2014), Micallef et al (2017), Mishra and Singh (2020). Learning methods Huang and Moraga (2004), Chen et al (2005), Mao et al (2006), Li and Liu (2009), Li, Chang, et al (2012), Guo et al (2017), Luo and Paal (2021).…”

Feature Extraction and Classification of High Dimensional Biomedical Spectral Data

“…자기공명 스펙트럼 분석기에 의해서 취득된 MRS (magnetic resonance spectra) 생체신호는 높은 입력차원 을 가지고 있을 뿐만 아니라, 데이터들간 유사성이 매우 높 기 때문에 패턴을 분류하는데 있어서 많은 어려움을 가지고 있다 [1]. 또한, 입력차원이 높은 신호를 데이터의 축소 없이 패턴을 분류할 경우 데이터 저장용량의 증가 및 많은 계산 처리에 의한 속도 저하를 초래한다.…”

A rapid screening test to distinguish betweenCandida albicansandCandida dubliniensisusing NMR spectroscopy

“…Thus simple, reliable, and rapid screening methods are still in demand to differentiate between C. dubliniensis and C. albicans. We have utilised NMR spectroscopy, combined with classification based on a statistical classification strategy (SCS), for microbial identification and developed a database to identify related, clinically relevant yeast species [11,12]. FTIR spectroscopy was used for identification of C. dubliniensis [8,10].…”

“…FTIR spectroscopy was used for identification of C. dubliniensis [8,10]. We have utilised NMR spectroscopy, combined with classification based on a statistical classification strategy (SCS), for microbial identification and developed a database to identify related, clinically relevant yeast species [11,12]. However, only distantly related species have been investigated.…”