Adoption of artificial intelligence (AI) by the medical community has long been anticipated, endorsed by a stream of machine learning literature showcasing AI systems that yield extraordinary performance. However, many of these systems are likely over-promising and will under-deliver in practice. One key reason is the community’s failure to acknowledge and address the presence of inflationary effects in the data. These simultaneously inflate evaluation performance and prevent a model from learning the underlying task, thus severely misrepresenting how that model would perform in the real world. This paper investigated the impact of these inflationary effects on healthcare tasks, as well as how these effects can be addressed. Specifically, we defined three inflationary effects that occur in medical data sets and allow models to easily reach small training losses and prevent skillful learning. We investigated two data sets of sustained vowel phonation from participants with and without Parkinson’s disease, and revealed that published models which have achieved high classification performances on these were artificially enhanced due to the inflationary effects. Our experiments showed that removing each inflationary effect corresponded with a decrease in classification accuracy, and that removing all inflationary effects reduced the evaluated performance by up to 30 percent. Additionally, the performance on a more realistic test set increased, suggesting that the removal of these inflationary effects enabled the model to better learn the underlying task and generalize. Source code is available at https://github.com/Wenbo-G/pd-phonation-analysis under the MIT license.

Resting-state electroencephalography (RSEEG) is a method under consideration as a potential biomarker that could support early and accurate diagnosis of Parkinson’s disease (PD). RSEEG data is often contaminated by signals arising from other electrophysiological sources and the environment, necessitating pre-processing of the data prior to applying machine learning methods for classification. Importantly, using differing degrees of pre-processing will lead to different classification results. This study aimed to examine this by evaluating the difference in experimental results when using re-referenced data, data that had undergone filtering and artefact rejection, and data without muscle artefact. The results demonstrated that, using a Random Forest Classifier for feature selection and a Support Vector Machine for disease classification, different levels of pre-processing led to markedly different classification results. In particular, the presence of muscle artefact was associated with inflated classification accuracy, emphasising the importance of its removal as part of pre-processing.

Parkinson’s Disease (PD) is a progressive chronic disorder with a high misdiagnosis rate. Because finger-tapping tasks correlate with its fine-motor symptoms, they could be used to help diagnose and assess PD. We first designed and developed an Android application to perform finger-tapping tasks without trained supervision, which is not always feasible for patients. Then, we conducted a preliminary user evaluation in Australia with six patients clinically diagnosed with PD and sixteen controls without PD. The application could be used in research and healthcare for regular symptom and progression assessment and feedback.

Current tests of disease status in Parkinson’s disease suffer from high variability, limiting their ability to determine disease severity and prognosis. Event-related potentials, in conjunction with machine learning, may provide a more objective assessment. In this study, we will use event-related potentials to develop machine learning models, aiming to provide an objective way to assess disease status and predict disease progression in Parkinson’s disease.

Computer-assisted quantification and analysis of postural sway may support identifying individuals affected by Parkinson’s disease (PD). Balancing, and its associated postural sway, is a complex process that requires the cooperation of several sensory systems in the brain. Unsurprisingly, a neurodegenerative disease can affect such processes, manifesting itself in the postural sway of affected individuals. Different aspects of postural sway can be quantified and represented as features, which can be used to distinguish between patients and controls. Our aim, inspired by a recent systematic literature review, was to experimentally determine whether sampling frequency and visual state had a meaningful impact on the effectiveness of features in distinguishing between the two groups, and whether overall discriminability could be improved using machine learning. We extracted 102 unique features from 78 postural sway recordings and found that the effectiveness (quantified by an effect size and the average area under the receiver operating characteristic curve) with a sampling frequency of 10 Hz was superior to 20, 40, and 100 Hz, though not with high confidence (quantified through Bayesian analysis). We also concluded that effectiveness under the eyes closed condition was higher than the eyes open condition (confirmed through Bayesian analysis), though combining features from both conditions was superior. Finally, we showed that using machine learning to analyse multiple features through feature selection resulted in higher discriminability in almost all cases. The code for these experiments have been released at https://github.com/Wenbo-G/pd-sway-analysis under the MIT license. When using our code, please cite this paper.

We measured postural sway in individuals diagnosed with Parkinson’s disease and age-matched controls. Individuals with Parkinson’s swayed more, as expected, especially when their eyes were closed. In the people with Parkinson’s, sway correlated strongly with cognitive measures, as well as with measures of quality of life and clinical status.

Background: Postural sway may be useful as an objective measure of Parkinson’s disease (PD). Existing studies have analyzed many different features of sway using different experimental paradigms. We aimed to determine what features have been used to measure sway and then to assess which feature(s) best differentiate PD patients from controls. We also aimed to determine whether any refinements might improve discriminative power and so assist in standardizing experimental conditions and analysis of data.

Methods: In this systematic review of the literature, effect size (ES) was calculated for every feature reported by each article and then collapsed across articles where appropriate. The influence of clinical medication status, visual state, and sampling rate on ES was also assessed. Results: Four hundred and forty‐three papers were retrieved. 25 contained enough information for further analysis. The most commonly used features were not the most effective (e.g., PathLength, used 14 times, had ES of 0.47, while TotalEnergy, used only once, had ES of 1.78). Increased sampling rate was associated with increased ES (PathLength ES increased to 1.12 at 100 Hz from 0.40 at 10 Hz). Measurement during “OFF” clinical status was associated with increased ES (PathLength ES was 0.83 OFF compared to 0.21 ON). Conclusions: This review identified promising features for analysis of postural sway in PD, recommending a sampling rate of 100 Hz and studying patients when OFF to maximize ES. ES complements statistical significance as it is clinically relevant and is easily compared across experiments. We suggest that machine learning is a promising tool for the future analysis of postural sway in PD.