Category: Dystonia: Pathophysiology, Imaging
Objective: To develop and validate an objective measure of the botulinum toxin treatment outcome in isolated dystonia.
Background: Dystonia is characterized by abnormal, often painful, postures and repetitive movements due to involuntary sustained or intermittent muscle contractions. Despite its debilitating impact on patients’ quality of life, clinical management of dystonia represents a significant challenge, with about one-third of patients not receiving any treatment. Botulinum neurotoxin (BoTX) injections into the affected muscles are considered the gold standard therapy for patients with focal and segmental dystonias. However, the variable effectiveness of BoTX therapy necessitates a repeated series of injections to refine the injection location, dosage, and administration regimen before the final treatment efficacy is established. An estimated 40% of patients with dystonia fail to benefit from BoTX treatment. One of the major factors limiting the efficient use of BoTX in dystonia patients is the absence of an objective predictive marker of treatment response.
Method: We developed and tested a deep learning platform, DystoniaBoTXNet, to predict the BoTX treatment outcome in 285 patients with focal dystonia (cervical dystonia, blepharospasm, laryngeal dystonia, writer’s cramp) based on their structural brain MRI and demographic information.
Results: The training model of DystoniaBoTXNet achieved an area under the curve (AUC) of 100% in discriminating 165 patients with or without BoTX treatment benefits based on a fully automated, data-driven identification of a neural biomarker of BoTX efficacy. In the independent testing sets, DystoniaBoTXNet achieved high accuracy in predicting the BoTX treatment outcome, including 92.9% accuracy in blepharospasm, 88.9% accuracy in cervical dystonia, 85% accuracy in laryngeal dystonia, and 76.9% accuracy in writer’s cramp.
Conclusion: The DystoniaBoTXNet deep learning platform is feasible for the objective and accurate algorithmic assessment of the BoTX treatment outcome prior to injection administration.
To cite this abstract in AMA style:
D. Yao, L. O'Flynn, K. Simonyan. DystoniaBoTXNet: A Deep Learning Platform for Predictive Outcome of Botulinum Toxin Treatment in Isolated Dystonia [abstract]. Mov Disord. 2022; 37 (suppl 2). https://www.mdsabstracts.org/abstract/dystoniabotxnet-a-deep-learning-platform-for-predictive-outcome-of-botulinum-toxin-treatment-in-isolated-dystonia/. Accessed September 7, 2024.« Back to 2022 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/dystoniabotxnet-a-deep-learning-platform-for-predictive-outcome-of-botulinum-toxin-treatment-in-isolated-dystonia/