Predicting impact of Deep Brain Stimulation on Non-motor symptoms of Parkinson's disease.

This is the largest study on Radiomics analysis looking into the impact of Deep Brain Stimulation on Non-Motor Symptoms (NMS) of Parkinson's disease. Preoperative brain white matter radiomics of 120 patients integrated with clinical variables were used to predict the DBS effect on NMS after 1 year from the surgery. Patients were classified "suboptimal" vs "good" based on a 10% or more improvement in NMS score. The combined Radiomics-Clinical Random Forrest (RF) model achieved an AUC of 0.96, Accuracy of 0.91, Sensitivity of 0.94 and Specificity of 0.88. The Youden's index showed optimal threshold for the RF of 0.535. The confusion matrix of the RF classifier gave a TPR of 0.92 and a FPR of 0.03. This corresponds to a PPV of 0.93 and a NPV of 0.93. The predictive models can be easily interpreted and after careful large-scale validation be integrated in assisting clinicians and patients to make informed decisions.Clinical Relevance- This paper shows the lesser studied positive impact of Deep Brain Stimulation on Non motor symptoms of Parkinson's disease while allows clinicians to predict non responders to the therapy.

Machine Learning Techniques for Personalized Detection of Epileptic Events in Clinical Video Recordings.

Continuous patient monitoring is essential to achieve an effective and optimal patient treatment in the intensive care unit. In the specific case of epilepsy it is the only way to achieve a correct diagnosis and a subsequent optimal medication plan if possible. In addition to automatic vital sign monitoring, epilepsy patients need manual monitoring by trained personnel, a task that is very difficult to be performed continuously for each patient. Moreover, epileptic manifestations are highly personalized even within the same type of epilepsy. In this work we assess two machine learning methods, dictionary learning and an autoencoder based on long short-term memory (LSTM) cells, on the task of personalized epileptic event detection in videos, with a set of features that were specifically developed with an emphasis on high motion sensitivity. According to the strengths of each method we have selected different types of epilepsy, one with convulsive behaviour and one with very subtle motion. The results on five clinical patients show a highly promising ability of both methods to detect the epileptic events as anomalies deviating from the stable/normal patient status.

Estimating normal and abnormal activities using smartphones.

The main objective of this study is to propose a computational pipeline for the recognition of normal and abnormal activities based on smartphone accelerometer data. Methods and techniques that have been previously evaluated are further evolved and applied for the recognition of a large set of separate activities as well as a sequence of activities simulating a common scenario of daily living as a more realistic approach. For these purposes, the MobiAct dataset which encompass a set of normal activities of daily living (ADLs) and abnormal activities (falls) was used. The results show a classification accuracy of 99% for the recognition of separate ADLs, while a reduction of 5% is observed for the recognition of the scenarios.

A computational platform for continuous seizure anticipation, monitoring and clinical evaluation.

The development of platforms that are able to continuously monitor and handle epileptic seizures in a non invasive manner is of great importance as they would improve the quality of life of drug resistant epileptic patients. In this work, a device and a computational platform is presented for acquiring low noise electroencephalographic signals, for the detection/prediction of epileptic seizures and the storage of ictal activity in an electronic personal health record. In order to develop this platform, a systematic clinical protocol was established including a number of drug resistant children from the University Hospital of Heraklion. Dry electrodes with innovative micro-spike design were proposed in order to increase the signal to noise ratio of the recorded EEG signals. A wearable low cost platform and its corresponding wireless communication protocol was developed focus on minimizing the interference with the patient's body. A computational subsystem with advanced algorithms provides detection/anticipation of upcoming seizure activity and aims to protect the patient from an accident due to a seizure or to improve his/her social life. Finally, the seizure activity information is stored in an electronic health record for further clinical evaluation.

An approach to absence epileptic seizures detection using Approximate Entropy.

Epilepsy is one of the most common chronic neurological diseases and the most common neurological chronic disease of childhood. The electroencephalogram (EEG) signal provides significant information neurologists take into consideration in the investigation and analysis of epileptic seizures. The Approximate Entropy (ApEn) is a formulated statistical parameter commonly used to quantify the regularity of a time series data of physiological signals. In this paper ApEn is used in order to detect the onset of epileptic seizures. The results show that the method provides promising results towards efficient detection of onset and ending of seizures, based on analyzing the corresponding EEG signals. ApEn parameters affect the method's behavior, suggesting that a more detailed study and a consistent methodology of their determination should be established. A preliminary analysis for the proper determination of these parameters is performed towards improving the results.

Absence seizure epilepsy detection using linear and nonlinear EEG analysis methods.

In this study, we investigated three measures capable of detecting absence seizures with increased sensitivity based on different underlying assumptions. Namely, an information-based method known as Approximate Entropy, a nonlinear alternative (Order Index), and a linear variance analysis approach. The results on the long-term EEG data suggest increased accuracy in absence seizure detection achieving sensitivity as high as 97.33% with no further application of any sophisticated classification scheme.

Vision-based motion detection, analysis and recognition of epileptic seizures--a systematic review.

The analysis of human motion from video has been the object of interest for many application areas, these including surveillance, control, biomedical analysis, video annotation etc. This paper addresses the advances within this topic in relation to epilepsy, a domain where human motion is with no doubt one of the most important elements of a patient's clinical image. It describes recent achievements in vision-based detection, analysis and recognition of human motion in epilepsy for marker-based and marker-free systems. An overview of motion-characterizing features extracted so far is presented separately. The objective is to gain existing knowledge in this field and set the route marks for the future development of an integrated decision support system for epilepsy diagnosis and disease management based on automated video analysis. This review revealed that the quantification of motion patterns of selected epileptic seizures has been studied thoroughly while the recognition of seizures is currently in its beginnings, but however feasible. Moreover, only a limited set of seizure types have been analyzed so far, indicating that a holistic approach addressing all epileptic syndromes is still missing.