Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning
<p>Whilst exaggerated bursts of beta frequency band oscillatory synchronization in the subthalamic nucleus (STN) have been associated with motor impairment in Parkinson’s disease (PD), a plausible mechanism linking the two phenomena has been lacking. Here I test the hypotheses that increased b...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
2021
|
| Subjects: |
| _version_ | 1797098133954494464 |
|---|---|
| author | Khawaldeh, SBK |
| author2 | Brown, P |
| author_facet | Brown, P Khawaldeh, SBK |
| author_sort | Khawaldeh, SBK |
| collection | OXFORD |
| description | <p>Whilst exaggerated bursts of beta frequency band oscillatory synchronization in the subthalamic nucleus (STN) have been associated with motor impairment in Parkinson’s disease (PD), a plausible mechanism linking the two phenomena has been lacking. Here I test the hypotheses that increased beta synchronization might compromise information coding capacity in basal ganglia (BG) networks, that beta activity is just one of several dynamic states in the STN local field potential (LFP) in PD, and that together these different states predict motor impairment with high fidelity. To this end, STN LFP activity was recorded in two experiments; the first in 18 PD patients as they executed cued upper and lower limb movements, and the second in 32 PD patients at rest following an overnight withdrawal of anti-parkinsonian medication, and after administration of levodopa.</p>
<p>In the first study, I used the accuracy of LFP-based limb-movement classification as an index of the information held by the STN. Machine learning (ML) using naïve Bayes showed that the intended movements could be predicted from the STN LFP recordings well ahead of their execution. The presence of bursts of LFP activity in the beta band significantly compromised the prediction of the limb to be moved, which denotes that exaggerated beta band synchronisation may restrict the capacity of the BG system to encode physiologically relevant information about intended actions.</p>
<p>In the second study, an unsupervised ML method based on hidden Markov modelling, was used to identify the nature and timing of transient states of distinct spectral content (e.g. theta, alpha, beta) in resting STN LFP recordings. This showed that levodopa reduced the occurrence rate and duration of low beta states, and the greater the reductions, the greater the improvement in motor impairment. However, additional LFP states were distinguished in the theta, alpha and high beta bands, and these behaved in an opposite manner. In addition, levodopa favoured the transition of low beta states to the other spectral states. When all LFP states and corresponding features were considered in a multivariate model it was possible to predict over 50% of the variance in patients’ hemibody impairment OFF medication, and in the change in hemibody impairment following levodopa.</p>
<p>These findings are important as they suggest that LFP activity may potentially be decoded to enable effector selection, in addition to force control in restorative brain-machine interface applications. Moreover, multiple spectral states in the STN LFP have a bearing on motor impairment, and that levodopa-induced shifts in the balance between these states can predict clinical change with high fidelity. This is important in suggesting that some states might be upregulated to improve parkinsonism and in suggesting how LFP feedback can be made more informative in closed-loop deep brain stimulation systems.</p> |
| first_indexed | 2024-03-07T05:05:18Z |
| format | Thesis |
| id | oxford-uuid:d9b976a9-a073-4b57-bb0b-432cca4218ad |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T05:05:18Z |
| publishDate | 2021 |
| record_format | dspace |
| spelling | oxford-uuid:d9b976a9-a073-4b57-bb0b-432cca4218ad2022-03-27T08:58:00ZSubthalamic nucleus bursting dynamics and prediction of motor measures using machine learningThesishttp://purl.org/coar/resource_type/c_db06uuid:d9b976a9-a073-4b57-bb0b-432cca4218adApplied Machine LearningBrain-computer interfacesProbabilistic Bayesian LearningParkinson's diseaseComputational NeuroscienceEnglishHyrax Deposit2021Khawaldeh, SBKBrown, PWoolrich, M<p>Whilst exaggerated bursts of beta frequency band oscillatory synchronization in the subthalamic nucleus (STN) have been associated with motor impairment in Parkinson’s disease (PD), a plausible mechanism linking the two phenomena has been lacking. Here I test the hypotheses that increased beta synchronization might compromise information coding capacity in basal ganglia (BG) networks, that beta activity is just one of several dynamic states in the STN local field potential (LFP) in PD, and that together these different states predict motor impairment with high fidelity. To this end, STN LFP activity was recorded in two experiments; the first in 18 PD patients as they executed cued upper and lower limb movements, and the second in 32 PD patients at rest following an overnight withdrawal of anti-parkinsonian medication, and after administration of levodopa.</p> <p>In the first study, I used the accuracy of LFP-based limb-movement classification as an index of the information held by the STN. Machine learning (ML) using naïve Bayes showed that the intended movements could be predicted from the STN LFP recordings well ahead of their execution. The presence of bursts of LFP activity in the beta band significantly compromised the prediction of the limb to be moved, which denotes that exaggerated beta band synchronisation may restrict the capacity of the BG system to encode physiologically relevant information about intended actions.</p> <p>In the second study, an unsupervised ML method based on hidden Markov modelling, was used to identify the nature and timing of transient states of distinct spectral content (e.g. theta, alpha, beta) in resting STN LFP recordings. This showed that levodopa reduced the occurrence rate and duration of low beta states, and the greater the reductions, the greater the improvement in motor impairment. However, additional LFP states were distinguished in the theta, alpha and high beta bands, and these behaved in an opposite manner. In addition, levodopa favoured the transition of low beta states to the other spectral states. When all LFP states and corresponding features were considered in a multivariate model it was possible to predict over 50% of the variance in patients’ hemibody impairment OFF medication, and in the change in hemibody impairment following levodopa.</p> <p>These findings are important as they suggest that LFP activity may potentially be decoded to enable effector selection, in addition to force control in restorative brain-machine interface applications. Moreover, multiple spectral states in the STN LFP have a bearing on motor impairment, and that levodopa-induced shifts in the balance between these states can predict clinical change with high fidelity. This is important in suggesting that some states might be upregulated to improve parkinsonism and in suggesting how LFP feedback can be made more informative in closed-loop deep brain stimulation systems.</p> |
| spellingShingle | Applied Machine Learning Brain-computer interfaces Probabilistic Bayesian Learning Parkinson's disease Computational Neuroscience Khawaldeh, SBK Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| title | Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| title_full | Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| title_fullStr | Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| title_full_unstemmed | Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| title_short | Subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| title_sort | subthalamic nucleus bursting dynamics and prediction of motor measures using machine learning |
| topic | Applied Machine Learning Brain-computer interfaces Probabilistic Bayesian Learning Parkinson's disease Computational Neuroscience |
| work_keys_str_mv | AT khawaldehsbk subthalamicnucleusburstingdynamicsandpredictionofmotormeasuresusingmachinelearning |