Machine learning prediction of motor response after deep brain stimulation in Parkinson's disease-proof of principle in a retrospective cohort

Aging suppresses subthalamic neuronal activity in patients with Parkinson's disease

Age is a primary risk factor for Parkinson's disease (PD); however, the effects of aging on the Parkinsonian brain remain poorly understood, particularly for deep brain structures. We investigated intraoperative micro-electrode recordings from the subthalamic nucleus (STN) of PD patients aged between 42 and 76 years. Age was associated with decreased oscillatory beta power and non-oscillatory high-frequency power, independent of PD-related variables. Single unit firing and burst rates were also reduced, whereas the coefficient of variation and the structure of burst activity were unchanged. Phase synchronization (debiased weighed phase lag index [dWPLI]) between sites was pronounced in the beta band between electrodes in the superficial STN but was unaffected by age. Our results show that aging is associated with reduced neuronal activity without changes to its temporal structure. We speculate that the loss of activity in the STN may mediate the relationship between PD and age.

Predicting tremor improvement after MRgFUS thalamotomy in essential tremor from preoperative spontaneous brain activity: A machine learning approach

Magnetic resonance-guided focused ultrasound surgery (MRgFUS) thalamotomy is an emerging technique for medication-refractory essential tremor (ET), but with variable outcomes. This study used pattern regression analysis to identify brain signatures predictive of tremor improvements. Fifty-four ET patients (mean age = 63.06 years, standard deviation (SD) = 10.55 years, 38 males) underwent unilateral MRgFUS thalamotomy and were scanned for resting-state functional magnetic resonance imaging (rs-fMRI). Seventy-four healthy controls (mean age = 58.09 years, SD = 10.30 years, 38 males) were recruited for comparison. Tremor responses at 12 months posttreatment were evaluated by the Clinical Rating Scale for Tremor. The fractional amplitude of low-frequency fluctuations (fALFF) was calculated from rs-fMRI data. Two-sample t-test was used to generate a disease-specific mask, within which Multivariate Kernel Ridge Regression analyses were conducted. Predicted and actual clinical scores were compared using Pearson's correlation coefficient (r) and normalized mean squared error (Norm. MSE). Permutation test and leave-one-out strategy were applied for results validation. KRR identified fALFF patterns that significantly predicted the hand tremor improvement (r = 0.23, P = 0.025; Norm. MSE = 0.05, P = 0.026) and the postural tremor improvement (r = 0.28, P = 0.025; Norm. MSE = 0.06, P = 0.023), but not action tremor improvement. Lobule VI of right cerebellum (Cerebelum_6_R), right superior occipital gyrus (Occipital_Sup_R) and lobule X of vermis (Vermis_10) contributed most for hand tremor prediction (normalized weights (NW): 2.77%, 2.40%, 2.34%) while Vermis_10, left supplementary motor area (Supp_Motor_Area_L) and right hippocampus (Hippocampus_R) for postural tremor prediction (NW: 2.69%, 2.12%, 2.05%). The low contributing NW of the individual brain regions suggested that the fALFF pattern as a whole is an overall predicting feature. Preoperative fALFF pattern predicts tremor benefits induced by MRgFUS thalamotomy. ClinicalTrials.gov number: NCT04570046.

Predictive modeling of sensory responses in deep brain stimulation

Introduction

Although stimulation-induced sensations are typically considered undesirable side effects in clinical DBS therapy, there are emerging scenarios, such as computer-brain interface applications, where these sensations may be intentionally created. The selection of stimulation parameters, whether to avoid or induce sensations, is a challenging task due to the vast parameter space involved. This study aims to streamline DBS parameter selection by employing a machine learning model to predict the occurrence and somatic location of paresthesias in response to thalamic DBS.

Methods

We used a dataset comprising 3,359 paresthetic sensations collected from 18 thalamic DBS leads from 10 individuals in two clinical centers. For each stimulation, we modeled the Volume of Tissue Activation (VTA). We then used the stimulation parameters and the VTA information to train a machine learning model to predict the occurrence of sensations and their corresponding somatic areas.

Results

Our results show fair to substantial agreement with ground truth in predicting the presence and somatic location of DBS-evoked paresthesias, with Kappa values ranging from 0.31 to 0.72. We observed comparable performance in predicting the presence of paresthesias for both seen and unseen cases (Kappa 0.72 vs. 0.60). However, Kappa agreement for predicting specific somatic locations was significantly lower for unseen cases (0.53 vs. 0.31).

Conclusion

The results suggest that machine learning can potentially be used to optimize DBS parameter selection, leading to faster and more efficient postoperative management. Outcome predictions may be used to guide clinical DBS programming or tuning of DBS based computer-brain interfaces.

Understanding societal challenges: a NeurotechEU perspective

Futuristic universities like The NeurotechEU and the technological innovations they provide will shape and serve society, but will also require support from society. Positive attitudes about neuro-technologies will increase their reach within society and may also impact policy-making, including funding decisions. However, the acceptability rates, especially of invasive neuro-technologies, are quite low and the majority of people are more worried than enthusiastic about them. The question therefore arises as to what neuro-technological advances should entail. In a rare effort to reach out to the public, we propose to conduct a trans-national survey with the goal to better understand the challenges of our NeurotechEU nations. We aim to compare and contrast our nations specifically with respect to their perspectives on neuro-technological advances, i.e., their needs for, interests in, access to, knowledge of and trust in neuro-technologies, and whether these should be regulated. To this end, we have developed the first version of a new tool-the Understanding Societal Challenges Questionnaire (USCQ)-which assesses all six of these dimensions (needs, interest, access, knowledge, trust, and policy-making) and is designed for administration across EU/AC countries. In addition to trans-national comparisons, we will also examine the links of our nations' perspectives on neuro-technological advances to demographic and personality variables, for example, education and socio-economic status, size of the residential area, the Big Five personality traits, religiosity, political standings, and more. We expect that this research will provide a deeper understanding of the challenges that our nations are facing as well as the similarities and differences between them, and will also help uncover the variables that predict positive and negative attitudes toward neuro-technological advances. By integrating this knowledge into the scientific process, The NeurotechEU may be able to develop neuro-technologies that people really care about, are ethical and regulated, and actually understood by the user.

Personalizing Deep Brain Stimulation Therapy for Parkinson's Disease With Whole-Brain MRI Radiomics and Machine Learning

Background Deep brain stimulation (DBS) is a well-recognised treatment for advanced Parkinson's disease (PD) patients. Structural brain alterations of the white matter can correlate with disease progression and act as a biomarker for DBS therapy outcomes. This study aims to develop a machine learning-driven predictive model for DBS patient selection using whole-brain white matter radiomics and common clinical variables. Methodology A total of 120 PD patients underwent DBS of the subthalamic nucleus. Their therapy effect was assessed at the one-year follow-up with the Unified Parkinson's Disease Rating Scale-part III (UPDRSIII) motor component. Radiomics analysis of whole-brain white matter was performed with PyRadiomics. The following machine learning methods were used: logistic regression (LR), support vector machine, naïve Bayes, K-nearest neighbours, and random forest (RF) to allow prediction of clinically meaningful UPRDSIII motor response before and after. Clinical variables were also added to the model to improve accuracy. Results The RF model showed the best performance on the final whole dataset with an area under the curve (AUC) of 0.99, accuracy of 0.95, sensitivity of 0.93, and specificity of 0.97. At the same time, the LR model showed an AUC of 0.93, accuracy of 0.88, sensitivity of 0.84, and specificity of 0.91. Conclusions Machine learning models can be used in clinical decision support tools which can deliver true personalised therapy recommendations for PD patients. Clinicians and engineers should choose between best-performing, less interpretable models vs. most interpretable, lesser-performing models. Larger clinical trials would allow to build trust among clinicians and patients to widely use these AI tools in the future.

Machine Learning in Neurosurgery: Toward Complex Inputs, Actionable Predictions, and Generalizable Translations

Machine learning can predict neurosurgical diagnosis and outcomes, power imaging analysis, and perform robotic navigation and tumor labeling. State-of-the-art models can reconstruct and generate images, predict surgical events from video, and assist in intraoperative decision-making. In this review, we will detail the neurosurgical applications of machine learning, ranging from simple to advanced models, and their potential to transform patient care. As machine learning techniques, outputs, and methods become increasingly complex, their performance is often more impactful yet increasingly difficult to evaluate. We aim to introduce these advancements to the neurosurgical audience while suggesting major potential roadblocks to their safe and effective translation. Unlike the previous generation of machine learning in neurosurgery, the safe translation of recent advancements will be contingent on neurosurgeons' involvement in model development and validation.

The Role of Levodopa Challenge in Predicting the Outcome of Subthalamic Deep Brain Stimulation

Background

Deep brain stimulation of the subthalamic nucleus (STN-DBS) is an effective and evidence-based treatment for idiopathic Parkinson's disease (iPD). A minority of patients does not sufficiently benefit from STN-DBS.

Objective

The predictive validity of the levodopa challenge for individual patients is analyzed.

Methods

Data from patients assessed with a preoperative Levodopa-test and a follow-up examination (mean ± standard deviation: 9.15 months ±3.39) from Kiel (n = 253), Berlin (n = 78) and Toronto (n = 98) were studied. Insufficient DBS outcome was defined as an overall UPDRS-III reduction <33% compared to UPDRS-III in med-off at baseline or alternatively if the minimal clinically important improvement of 5 points was not reached. Single UPDRS-items and sub-scores were dichotomized. Following exploratory analysis, we trained supervised regression- and classification models for outcome prediction.

Results

Data analysis confirmed significant correlation between the absolute UPDRS-III reduction during Levodopa challenge and after stimulation. But individual improvement was inaccurately predicted with a large range of up to 30 UPDRS III points. Further analysis identified preoperative UPDRS-III/med-off-scores and preoperative Levodopa-improvement as most influential factors. The models for UPDRS-III and sub-scores improvement achieved comparably low accuracy.

Conclusions

With large prediction intervals, the Levodopa challenge use for patient counseling is limited, though remains important for excluding non-responders to Levodopa. Despite these deficiencies, the current practice of patient selection is highly successful and builds not only on the Levodopa challenge. However, more specific motor tasks and further paraclinical tools for prediction need to be developed.

The altered spontaneous neural activity in patients with Parkinson's disease and its predictive value for the motor improvement of deep brain stimulation

BACKGROUND

This study aims to investigate the altered spontaneous neural activity in patients with Parkinson's disease (PD) revealed by amplitudes of low-frequency fluctuations (ALFF) of resting-state fMRI, and the feasibility of using ALFF as neuroimaging predictors for motor improvement after bilateral subthalamic nucleus (STN) deep brain stimulation (DBS).

METHODS

Fourty-four patients and 44 healthy controls were included in this study. First, the ALFF of patients with PD was compared with that of controls; then significant clusters were correlated with motor improvement after DBS (unified Parkinson's disease rating scale (UPDRS-III)) and other clinical variables. Second, regression and classification of the machine learning models were conducted to predict motor improvement after DBS. Receiver operating characteristic (ROC) analysis was used to evaluate the performance of the classification model.

RESULTS

Compared with healthy controls, patients with PD showed increased ALFF in the bilateral motor area and decreased ALFF in the bilateral temporal cortex and cerebellum. The Hoehn-Yahr stages correlated with ALFF within the bilateral cerebellum (p = 0.021), and UPDRS-III improvement correlated with ALFF in the left (p < 0.001) and right (p = 0.005) motor areas. The regression model showed a significant correlation between the predicted and observed UPDRS-III changes (R = 0.65, p < 0.001). The ROC analysis revealed an area under the curve (AUC) of 0.94 which differentiated moderate and superior DBS responders.

CONCLUSION

The results revealed altered ALFF patterns in patients with PD and their correlations with clinical variables. Both binary and continuous ALFF can potentially serve as predictive biomarkers for DBS response.

Discovering Themes in Deep Brain Stimulation Research Using Explainable Artificial Intelligence

Deep brain stimulation is a treatment that controls symptoms by changing brain activity. The complexity of how to best treat brain dysfunction with deep brain stimulation has spawned research into artificial intelligence approaches. Machine learning is a subset of artificial intelligence that uses computers to learn patterns in data and has many healthcare applications, such as an aid in diagnosis, personalized medicine, and clinical decision support. Yet, how machine learning models make decisions is often opaque. The spirit of explainable artificial intelligence is to use machine learning models that produce interpretable solutions. Here, we use topic modeling to synthesize recent literature on explainable artificial intelligence approaches to extracting domain knowledge from machine learning models relevant to deep brain stimulation. The results show that patient classification (i.e., diagnostic models, precision medicine) is the most common problem in deep brain stimulation studies that employ explainable artificial intelligence. Other topics concern attempts to optimize stimulation strategies and the importance of explainable methods. Overall, this review supports the potential for artificial intelligence to revolutionize deep brain stimulation by personalizing stimulation protocols and adapting stimulation in real time.

Developments in the mechanistic understanding and clinical application of deep brain stimulation for Parkinson's disease

INTRODUCTION. Deep brain stimulation (DBS) is a life-changing treatment for patients with Parkinson's disease (PD) and gives the unique opportunity to directly explore how basal ganglia work. Despite the rapid technological innovation of the last years, the untapped potential of DBS is still high.

AREAS COVERED. This review summarizes the developments in the mechanistic understanding of DBS and the potential clinical applications of cutting-edge technological advances. Rather than a univocal local mechanism, DBS exerts its therapeutic effects through several multimodal mechanisms and involving both local and network-wide structures, although crucial questions remain unexplained. Nonetheless, new insights in mechanistic understanding of DBS in PD have provided solid bases for advances in preoperative selection phase, prediction of motor and non-motor outcomes, leads placement and postoperative stimulation programming.

EXPERT OPINION. DBS has not only strong evidence of clinical effectiveness in PD treatment, but technological advancements are revamping its role of neuromodulation of brain circuits and key to better understanding PD pathophysiology. In the next few years, the worldwide use of new technologies in clinical practice will provide large data to elucidate their role and to expand their applications for PD patients, providing useful insights to personalize DBS treatment and follow-up.

Proceedings of the Ninth Annual Deep Brain Stimulation Think Tank: Advances in Cutting Edge Technologies, Artificial Intelligence, Neuromodulation, Neuroethics, Pain, Interventional Psychiatry, Epilepsy, and Traumatic Brain Injury

DBS Think Tank IX was held on August 25-27, 2021 in Orlando FL with US based participants largely in person and overseas participants joining by video conferencing technology. The DBS Think Tank was founded in 2012 and provides an open platform where clinicians, engineers and researchers (from industry and academia) can freely discuss current and emerging deep brain stimulation (DBS) technologies as well as the logistical and ethical issues facing the field. The consensus among the DBS Think Tank IX speakers was that DBS expanded in its scope and has been applied to multiple brain disorders in an effort to modulate neural circuitry. After collectively sharing our experiences, it was estimated that globally more than 230,000 DBS devices have been implanted for neurological and neuropsychiatric disorders. As such, this year's meeting was focused on advances in the following areas: neuromodulation in Europe, Asia and Australia; cutting-edge technologies, neuroethics, interventional psychiatry, adaptive DBS, neuromodulation for pain, network neuromodulation for epilepsy and neuromodulation for traumatic brain injury.

Predicting Motor Responsiveness to Deep Brain Stimulation with Machine Learning

Deep brain stimulation is a complex movement disorder intervention that requires highly invasive brain surgery. Clinicians struggle to predict how patients will respond to this treatment. To address this problem, we are working toward developing a clinical tool to help neurologists predict deep brain stimulation response. We analyzed a cohort of 105 Parkinson's patients who underwent deep brain stimulation at Vanderbilt University Medical Center. We developed binary and multicategory models for predicting likelihood of motor symptom reduction after undergoing deep brain stimulation. We compared the performances of our best models to predictions made by neurologist experts in movement disorders. The strongest binary classification model achieved a 10-fold cross validation AUC of 0.90, outperforming the best neurologist predictions (0.56). These results are promising for future clinical applications, though more work is necessary to validate these findings in a larger cohort and taking into consideration broader quality of life outcome measures.

Multicenter Validation of Individual Preoperative Motor Outcome Prediction for Deep Brain Stimulation in Parkinson's Disease

BACKGROUND

Subthalamic nucleus deep brain stimulation (STN DBS) is an established therapy for Parkinson's disease (PD) patients suffering from motor response fluctuations despite optimal medical treatment, or severe dopaminergic side effects. Despite careful clinical selection and surgical procedures, some patients do not benefit from STN DBS. Preoperative prediction models are suggested to better predict individual motor response after STN DBS. We validate a preregistered model, DBS-PREDICT, in an external multicenter validation cohort.

METHODS

DBS-PREDICT considered eleven, solely preoperative, clinical characteristics and applied a logistic regression to differentiate between weak and strong motor responders. Weak motor response was defined as no clinically relevant improvement on the Unified Parkinson's Disease Rating Scale (UPDRS) II, III, or IV, 1 year after surgery, defined as, respectively, 3, 5, and 3 points or more. Lower UPDRS III and IV scores and higher age at disease onset contributed most to weak response predictions. Individual predictions were compared with actual clinical outcomes.

RESULTS

322 PD patients treated with STN DBS from 6 different centers were included. DBS-PREDICT differentiated between weak and strong motor responders with an area under the receiver operator curve of 0.76 and an accuracy up to 77%.

CONCLUSION

Proving generalizability and feasibility of preoperative STN DBS outcome prediction in an external multicenter cohort is an important step in creating clinical impact in DBS with data-driven tools. Future prospective studies are required to overcome several inherent practical and statistical limitations of including clinical decision support systems in DBS care.

Prediction of Deep Brain Stimulation Outcome in Parkinson's Disease With Connectome Based on Hemispheric Asymmetry

Parkinson’s disease (PD) is a neurodegenerative disease that is associated with motor and non-motor symptoms and caused by lack of dopamine in the substantia nigra of the brain. Subthalamic nucleus deep brain stimulation (STN-DBS) is a widely accepted therapy of PD that mainly inserts electrodes into both sides of the brain. The effect of STN-DBS was mainly for motor function, so this study focused on the recovery of motor function for PD after DBS. Hemispherical asymmetry in the brain network is considered to be a potential indicator for diagnosing PD patients. This study investigated the value of hemispheric brain connection asymmetry in predicting the DBS surgery outcome in PD patients. Four types of brain connections, including left intra-hemispheric (LH) connection, right intra-hemispheric (RH) connection, inter-hemispheric homotopic (Ho) connection, and inter-hemispheric heterotopic (He) connection, were constructed based on the resting state functional magnetic resonance imaging (rs-fMRI) performed before the DBS surgery. We used random forest for selecting features and the Ridge model for predicting surgical outcome (i.e., improvement rate of motor function). The functional connectivity analysis showed that the brain has a right laterality: the RH networks has the best correlation (r = 0.37, p = 5.68E-03) between the predicted value and the true value among the above four connections. Moreover, the region-of-interest (ROI) analysis indicated that the medioventral occipital cortex (MVOcC)–superior temporal gyrus (STG) and thalamus (Tha)–precentral gyrus (PrG) contributed most to the outcome prediction model for DBS without medication. This result provides more support for PD patients to evaluate DBS before surgery.

Identifying the neurophysiological effects of memory-enhancing amygdala stimulation using interpretable machine learning

BACKGROUND

Direct electrical stimulation of the amygdala can enhance declarative memory for specific events. An unanswered question is what underlying neurophysiological changes are induced by amygdala stimulation.

OBJECTIVE

To leverage interpretable machine learning to identify the neurophysiological processes underlying amygdala-mediated memory, and to develop more efficient neuromodulation technologies.

METHOD

Patients with treatment-resistant epilepsy and depth electrodes placed in the hippocampus and amygdala performed a recognition memory task for neutral images of objects. During the encoding phase, 160 images were shown to patients. Half of the images were followed by brief low-amplitude amygdala stimulation. For local field potentials (LFPs) recorded from key medial temporal lobe structures, feature vectors were calculated by taking the average spectral power in canonical frequency bands, before and after stimulation, to train a logistic regression classification model with elastic net regularization to differentiate brain states.

RESULTS

Classifying the neural states at the time of encoding based on images subsequently remembered versus not-remembered showed that theta and slow-gamma power in the hippocampus were the most important features predicting subsequent memory performance. Classifying the post-image neural states at the time of encoding based on stimulated versus unstimulated trials showed that amygdala stimulation led to increased gamma power in the hippocampus.

CONCLUSION

Amygdala stimulation induced pro-memory states in the hippocampus to enhance subsequent memory performance. Interpretable machine learning provides an effective tool for investigating the neurophysiological effects of brain stimulation.

PassFlow: a multimodal workflow for predicting deep brain stimulation outcomes

PURPOSE

Deep Brain Stimulation (DBS) is a proven therapy for Parkinson's Disease (PD), frequently resulting in an enhancement of motor function. Nonetheless, several undesirable side effects can occur after DBS, which can worsen the quality of life of the patient. Thus, the clinical team has to carefully select patients on whom to perform DBS. Over the past decade, there have been some attempts to relate pre-operative data and DBS clinical outcomes, with most focused on the motor symptomatology. In this paper, we propose a machine learning-based method able to predict a large number of DBS clinical outcomes for PD.

METHODS

We propose a multimodal pipeline, referred to as PassFlow, which predicts 84 clinical post-operative clinical scores. PassFlow is composed of an artificial neural network to compress clinical information, an image processing method from the state-of-the-art to extract morphological biomarkers our of T1 imaging, and an SVM to perform the regressions. We validated PassFlow on 196 PD patients who undergone a DBS.

RESULTS

PassFlow showed correlation coefficients as high as 0.71 and were able to significantly predict 63 out of the 84 scores, outperforming a comparative linear method. The number of metrics that are predicted with this pre-operative information was also found to be correlated with the number of patients with this information available, indicating that the PassFlow method is still actively learning.

CONCLUSION

We presented a novel, machine learning-based pipeline to predict a variety of post-operative clinical outcomes of DBS for PD patients. PassFlow took into account various bio-markers, arising from different data modalities, showing high correlation coefficients for some scores from pre-operative data only. It indicates that many clinical outcomes of DBS can be predicted agnostic to the specific simulation parameters, as PassFlow has been validated without such stimulation-related information.

Machine learning-based personalized subthalamic biomarkers predict ON-OFF levodopa states in Parkinson patients

Objective.Adaptive deep brain stimulation (aDBS) based on subthalamic nucleus (STN) electrophysiology has recently been proposed to improve clinical outcomes of DBS for Parkinson's disease (PD) patients. Many current models for aDBS are based on one or two electrophysiological features of STN activity, such as beta or gamma activity. Although these models have shown interesting results, we hypothesized that an aDBS model that includes many STN activity parameters will yield better clinical results. The objective of this study was to investigate the most appropriate STN neurophysiological biomarkers, detectable over long periods of time, that can predict OFF and ON levodopa states in PD patients.Approach.Long-term local field potentials (LFPs) were recorded from eight STNs (four PD patients) during 92 recording sessions (44 OFF and 48 ON levodopa states), over a period of 3-12 months. Electrophysiological analysis included the power of frequency bands, band power ratio and burst features. A total of 140 engineered features was extracted for 20 040 epochs (each epoch lasting 5 s). Based on these engineered features, machine learning (ML) models classified LFPs as OFF vs ON levodopa states.Main results.Beta and gamma band activity alone poorly predicts OFF vs ON levodopa states, with an accuracy of 0.66 and 0.64, respectively. Group ML analysis slightly improved prediction rates, but personalized ML analysis, based on individualized engineered electrophysiological features, were markedly better, predicting OFF vs ON levodopa states with an accuracy of 0.8 for support vector machine learning models.Significance.We showed that individual patients have unique sets of STN neurophysiological biomarkers that can be detected over long periods of time. ML models revealed that personally classified engineered features most accurately predict OFF vs ON levodopa states. Future development of aDBS for PD patients might include personalized ML algorithms.