Discovering the Discriminating Power in Patient Test Features Using Visual Analytics: A Case Study in Parkinson’s Disease

Abstract

Part 11: New Methods and Tools for Big Data Wokshop (MT4BD)International audienceThis paper presents a novel methodology for selecting the most representative features for identifying the presence of the Parkinson’s Disease (PD). The proposed methodology is based on interactive visual analytic based on multi-objective optimisation. The implemented tool processes and visualises the information extracted via performing a typical line-tracking test using a tablet device. Such output information includes several modalities, such as position, velocity, dynamics, etc. Preliminary results depict that the implemented visual analytics technique has a very high potential in discriminating the PD patients from healthy individuals and thus, it can be used for the identification of the best feature type which is representative of the disease presence

Explaining artificial intelligence with visual analytics in healthcare

“…Examples of modifiable aspects are: the parameter k in k-nearest neighbors clustering (Spitz et al, 2020), the number of clusters (X. Ji, Shen, et al, 2019;Figure 5b), the applied algorithm (Riegler et al, 2016), the attributes for analysis (Abbasloo et al, 2019;Dixit et al, 2017;Zhao et al, 2017), the query for pattern mining (Gotz et al, 2014;Figure 2e), and the feature weights that best distinguish ill and healthy people (Moschonas et al, 2016).…”

Novel ICT biomedical applications towards precision medicine of chronic diseases

“…The current methodology relies on the results of the previous study of [Mos+16]. The Line Test methodology was implemented in a group of participants.…”

“…To analyze the complicated results produced by the Line Test, the multi-objective visualization of [Mos+16] was modified based on the current features. The results of each patient were used in order to select the most significant through feature extraction methods.…”

“…A novel methodology able to identify the presence of the Parkinson's Disease (PD) based on interactive visual analytic based on multi-objective optimization was utilized within NoTremor project [Mos+16]. A typical motion tracking methodology is performed to a study cohort and a set of output information includes several modalities, such as position, velocity, dynamics, etc are processed by the implemented tool.…”

“…The aim of this work is to demonstrate how a computer based motion monitoring method can be utilized for patient classification and assess the severity of PD. The real-time and reliable analysis of patient's motor abilities was established in our previous work [Mos+16]. The electronic device translates patient's move into a set of features that depicts the movement's main characteristics.…”