Application of Machine Learning Using Decision Trees for Prognosis of Deep Brain Stimulation of Globus Pallidus Internus for Children With Dystonia

Abstract

Background: While Deep Brain Stimulation (DBS) of the Globus pallidus internus is a well-established therapy for idiopathic/genetic dystonia, benefits for acquired dystonia are varied, ranging from modest improvement to deterioration. Predictive biomarkers to aid DBS prognosis for children are lacking, especially in acquired dystonias, such as dystonic Cerebral Palsy. We explored the potential role of machine learning techniques to identify parameters that could help predict DBS outcome. Methods: We conducted a retrospective study of 244 children attending King's College Hospital between September 2007 and June 2018 for neurophysiological tests as part of their assessment for possible DBS at Evelina London Children's Hospital. For the 133 individuals who underwent DBS and had 1-year outcome data available, we assessed the potential predictive value of six patient parameters: sex, etiology (including cerebral palsy), baseline severity (Burke-Fahn-Marsden Dystonia Rating Scale-motor score), cranial MRI and two neurophysiological tests, Central Motor Conduction Time (CMCT) and Somatosensory Evoked Potential (SEP). We applied machine learning analysis to determine the best combination of these features to aid DBS prognosis. We developed a classification algorithm based on Decision Trees (DTs) with k-fold cross validation for independent testing. We analyzed all possible combinations of the six features and focused on acquired dystonias. Results: Several trees resulted in better accuracy than the majority class classifier. However, the two features that consistently appeared in top 10 DTs were CMCT and baseline dystonia severity. A decision tree based on CMCT and baseline severity provided a range of sensitivity and specificity, depending on the threshold chosen for baseline dystonia severity. In situations where CMCT was not available, a DT using SEP alone provided better than the majority class classifier accuracy. Shah et al. Childhood-Dystonia: Data-Driven DBS Prognosis Conclusion: The results suggest that neurophysiological parameters can help predict DBS outcomes, and DTs provide a data-driven, highly interpretable decision support tool that lends itself to being used in clinical practice to help predict potential benefit of DBS in dystonic children. Our results encourage the introduction of neurophysiological parameters in assessment pathways, and data collection to facilitate multi-center evaluation and validation of these potential predictive markers and of the illustrative decision support tools presented here.

Dystonia Management: What to Expect From the Future? The Perspectives of Patients and Clinicians Within DystoniaNet Europe

“…At a subsequent stage, it could be supplemented with infrastructures for biobanking. Indeed, although a growing amount of data suggest that some patient characteristics may inform patient selection for surgery (105)(106)(107)(108), at the moment, there are only tentative clinical, neuroradiological, genetic, or neurophysiological elements that could predict individual surgery outcome [as described above (66)]. Such much-needed biomarkers need to be rolled out across a wider population and pooled diagnostic subgroups, and this requires a collective effort, where different centers would not only contribute clinical data but also share infrastructures and expertise in the different fields.…”

“…Combining different techniques together, such as supervised machine learning applied to standard diagnostic brain MRI together with measuring central motor conduction times (CMCT) with transcranial magnetic stimulation (TMS), or -evoked potentials (SEPs) together with dystonia severity scales, can help counsel patients and families of dystonic children regarding the likely benefit of DBS in acquired dystonias as well as provide personal predictive and decision-making data using receiver operating characteristic (ROC) curves ( 66 ). This process applied internationally could rapidly build gene-specific and acquired disease-specific decision-making tools.…”

The Expanding Horizon of Neural Stimulation for Hyperkinetic Movement Disorders

“…The generation of response-led classifications is also an emergent theme and here the use of machine learning can guide development. Such an approach was recently used with good effect in childhood dystonia (56). In this study, six patient parameters (sex, etiology, baseline severity, cranial MRI and central motor conduction time and/or sensory evoked potential) were evaluated for their ability predict deep brain stimulation outcome using a decision tree supervised learning method (Figure 1A).…”

“…The middle node then examines whether the corticospinal tract is intact using abnormalities in the central motor conduction time (CMT) as the delineator. Finally, more severe disease is predictive of a good response [adapted from (56)]. (B) This panel exemplifies a real time closed loop algorithm which has been successfully used in essential tremor.…”

Machine Learning Models on ADC Features to Assess Brain Changes of Children With Pierre Robin Sequence

“…Machine learning, however, can uncover important information that is undetected by routine clinical medical imaging, thus revealing potential biological information (11)(12)(13). Given its compelling advantages in the implementation of classification and prediction (14,15), machine learning has become a common method of scientific research (16,17). In this study, we analyzed ADC features combined with several machine learning models to effectively evaluate brain development differences between children with PRs and normal controls as well as the validity of the model.…”