Ling Yun Yeow, Geetha Chilla, Chew Qian Hui, Sim Kang, Bhanu Prakash K.N. Psychometric scores-based feature identification and classification of Schizophrenics and healthy controls using Machine learning methods – An Explorative study. SingHealth Duke-NUS Scientific Congress 2021
Abstract:
Schizophrenia is a major psychotic illness which causes one to lose touch with reality. Assessment tools such as Positive and Negative Syndrome Scale (PANSS), brief form of World Health Organization Quality of Life (QOL), Wide Range Achievement Test (WRAT3) and Brief Assessment of Cognition in Schizophrenia (BACS) have been used in patients with Schizophrenia and healthy control cohorts. Initial data exploratory analysis using t-Distributed Stochastic Neighbor Embedding (t-SNE) with these psychometric scores showed some overlaps in both the cohorts. This prompted further investigation by utilizing machine learning (ML) methods to identify the more important psychometric features followed by classifying between SZ and HC and assessing the performances of the classifiers to determine how well we can use these psychometric scores to delineate SZ from HC. We demonstrated that reading ability (WRAT3), four domains of QOL and cognitive functioning (BACS) were important features. RF with tuned hyper-parameters outperformed other methods, thus giving proof of concept support for further use of functional and cognitive measures in differentiating between subtypes of the illness.
License type:
Publisher Copyright
Funding Info:
There was no specific funding for the research done