A MATLAB App to Assess, Compare and Validate New Methods Against Their Benchmarks
Emerging technologies for physiological signals and data collection enable the monitoring of patient health and well-being in real-life settings. This requires novel methods and tools to compare the validity of this kind of information with that acquired
- PDF / 2,829,711 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 84 Downloads / 197 Views
2
Applied Biomedical Signal Processing and Intelligent eHealth Lab, School of Engineering, University of Warwick, Coventry CV4 7AL, UK [email protected] Department of Mechanical and Aerospace Engineering, Sapienza University of Rome, 00184 Rome, Italy
Abstract. Emerging technologies for physiological signals and data collection enable the monitoring of patient health and well-being in real-life settings. This requires novel methods and tools to compare the validity of this kind of information with that acquired in controlled environments using more costly and sophisticated technologies. In this paper, we describe a method and a MATLAB tool that relies on a standard sequence of statistical tests to compare features obtained using novel techniques with those acquired by means of benchmark procedures. After introducing the key steps of the proposed statistical analysis method, this paper describes its implementation in a MATLAB app, developed to support researchers in testing the extent to which a set of features, captured with a new methodology, can be considered a valid surrogate of that acquired employing gold standard techniques. An example of the application of the tool is provided in order to validate the method and illustrate the graphical user interface (GUI). The app development in MATLAB aims to improve its accessibility, foster its rapid adoption among the scientific community and its scalability into wider MATLAB tools. Keywords: Surrogate features
Statistical analysis Benchmarking
1 Introduction Biomedical engineers, among other experts of Science, Technology, Engineering and Mathematics (STEM), constantly attempt to develop novel approaches and techniques to acquire and/or analyze different kinds of data [1–3]. As an example, the unprecedented amount of data generated by the spread use of Internet of Things (IoT) is encouraging the deployment of alternative methods and tools. Normally these alternative methods are less time-consuming, less expensive, non-invasive and can work automatically, with minimum manual intervention of an expert. This is a positive trend, which can, however, lead to the improper use and application of methodologies that are not yet correctly validated and whose results are consequently unreliable [4]. For instance, as widely discussed by Pecchia et al. [4], © Springer Nature Switzerland AG 2021 T. Jarm et al. (Eds.): EMBEC 2020, IFMBE Proceedings 80, pp. 10–21, 2021. https://doi.org/10.1007/978-3-030-64610-3_2
A MATLAB App to Assess, Compare and Validate New Methods
11
several studies employed only correlations to prove that ultra-short term heart rate variability (HRV) features behaved as short term ones, concluding that the former were good substitutes of the latter if significantly correlated among each other [5]. Nonetheless, this result is controversial, because “a correlate does not make a surrogate” [6]. Specifically, in [4], Pecchia et al. presented a protocol to evaluate whether ultrashort terms HRV features can be considered as a valid replacement for short
Data Loading...
