
Dr Chloe Hinchliffe
About
My research project
Application of Machine Learning for the Differential Diagnosis of Psychogenic Nonepileptic Seizures and EpilepsyPsychogenic non-epileptic seizures (PNES) are characterised by a lack of epileptic activity in the brain and about one in five referrals to epilepsy clinics actually have this condition. PNES is diagnosed by recording a seizure using video-electroencephalogram (EEG), from which an expert inspects the semiology and the EEG. This method is reliable but expensive, inconvenient for the patient, and not accessible for all hospitals. This could be improved with machine learning classifiers, which are models that can consider multiple inputs at once. Since no single biomarker has been found to diagnose PNES, these classifiers could be a very useful aid to clinicians.
The hypothesis of this PhD is that is it possible to identify subjects with PNES from those with epilepsy using non-ictal biomedical signals with machine learning. This was tested using a data set of interictal and preictal EEG and electrocardiograms (ECG) recordings from 48 subjects with PNES and 29 subjects with epilepsy. A wide range of features were extracted from the signals and grouped into 'families'. The performance of the different feature families was evaluated using statistical methods and ranked using two feature ranking methods. The families and the signals themselves were then classified using several machine learning models.
The highest classification accuracy reported from purely statistical analysis was 60.67\%. The highest balanced accuracy reported by the machine learning models, however, was 97.00\% from the 'all-reduced' family. Therefore, machine learning was much more effective than using individual features. This PhD has therefore shown that machine learning could be a powerful aid in PNES diagnosis, which can limit the costs by reducing the need for diagnosis with a video recording and review by an expert.
Supervisors
Psychogenic non-epileptic seizures (PNES) are characterised by a lack of epileptic activity in the brain and about one in five referrals to epilepsy clinics actually have this condition. PNES is diagnosed by recording a seizure using video-electroencephalogram (EEG), from which an expert inspects the semiology and the EEG. This method is reliable but expensive, inconvenient for the patient, and not accessible for all hospitals. This could be improved with machine learning classifiers, which are models that can consider multiple inputs at once. Since no single biomarker has been found to diagnose PNES, these classifiers could be a very useful aid to clinicians. The hypothesis of this PhD is that is it possible to identify subjects with PNES from those with epilepsy using non-ictal biomedical signals with machine learning. This was tested using a data set of interictal and preictal EEG and electrocardiograms (ECG) recordings from 48 subjects with PNES and 29 subjects with epilepsy. A wide range of features were extracted from the signals and grouped into 'families'. The performance of the different feature families was evaluated using statistical methods and ranked using two feature ranking methods. The families and the signals themselves were then classified using several machine learning models. The highest classification accuracy reported from purely statistical analysis was 60.67\%. The highest balanced accuracy reported by the machine learning models, however, was 97.00\% from the 'all-reduced' family. Therefore, machine learning was much more effective than using individual features. This PhD has therefore shown that machine learning could be a powerful aid in PNES diagnosis, which can limit the costs by reducing the need for diagnosis with a video recording and review by an expert.
Publications
Psychogenic non-epileptic seizures (PNES) may resemble epileptic seizures but are not caused by epileptic activity. However, the analysis of electroencephalogram (EEG) signals with entropy algorithms could help identify patterns that differentiate PNES and epilepsy. Furthermore, the use of machine learning could reduce the current diagnosis costs by automating classification. The current study extracted the approximate sample, spectral, singular value decomposition, and Renyi entropies from interictal EEGs and electrocardiograms (ECG)s of 48 PNES and 29 epilepsy subjects in the broad, delta, theta, alpha, beta, and gamma frequency bands. Each feature-band pair was classified by a support vector machine (SVM), k-nearest neighbour (kNN), random forest (RF), and gradient boosting machine (GBM). In most cases, the broad band returned higher accuracy, gamma returned the lowest, and combining the six bands together improved classifier performance. The Renyi entropy was the best feature and returned high accuracy in every band. The highest balanced accuracy, 95.03%, was obtained by the kNN with Renyi entropy and combining all bands except broad. This analysis showed that entropy measures can differentiate between interictal PNES and epilepsy with high accuracy, and improved performances indicate that combining bands is an effective improvement for diagnosing PNES from EEGs and ECGs.
Abstract - Psychogenic non-epileptic seizures (PNES) are attacks that resemble epilepsy but are not associated with epileptic brain activity and are regularly misdiagnosed. The current gold standard method of diagnosis is expensive and complex. Electroencephalogram (EEG) analysis with machine learning could improve this. A k-nearest neighbours (kNN) and support vector machine (SVM) were used to classify EEG connectivity measures from 48 patients with PNES and 29 patients with epilepsy. The synchronisation method-correlation or coherence-and the binarisation threshold were defined through experimentation. Ten network parameters were extracted from the synchronisation matrix. The broad, delta, theta, alpha, beta, gamma, and combined 'all' frequency bands were compared along with three feature selection methods: the full feature set (no selection), light gradient boosting machine (LGBM) and k-Best. Coherence was the highest performing synchronisation method and 0.6 was the best coherence threshold. The highest balanced accuracy was 89.74%, produced by combining all six frequency bands and selecting features with LGBM, classified by the SVM. This method returned a comparatively high accuracy but at a high computation cost. Future research should focus on identifying specific frequency bands and network parameters to reduce this cost. Clinical relevance - This study found that EEG connectivity and machine learning methods can be used to differentiate PNES from epilepsy using interictal recordings to a high accuracy. Thus, this method could be an effective tool in assisting clinicians in PNES diagnosis without a video-EEG recording of a habitual seizure.