Journal of neurosurgery · 2026
Comparative analysis of machine learning algorithms for predicting stereotactic coordinates of the centromedian nucleus.
СтатьяСравнительное исследованиеНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Singh Hargunbir, Bhatia Nimit, Shless Jared, Stamm Michaela, Warren Aaron E L, Kulsomphob Avelyn, Pacheco-Barrios Niels, Chua Melissa M J, Rolston John D
Аннотация
OBJECTIVE: Deep brain stimulation (DBS) of the centromedian nucleus (CM) of the thalamus is a promising treatment for drug-resistant epilepsy, Tourette syndrome, disorders of consciousness, and chronic pain, particularly when other surgical options are not feasible. However, the CM is challenging to visualize on routine MRI and atlas-based targeting often results in inaccurate electrode placement, affecting surgical outcomes. Furthermore, inability to visualize and directly target the CM is a barrier to CM-DBS in a resource-limited setting. The aim of this study was to develop and test machine learning (ML) models that could predict target coordinates of the CM using multiple datapoints available from conventional T1-weighted MRI.
METHODS: Four ML models-linear regression (LR), k-nearest neighbor (KNN), support vector regression (SVR), and deep neural network (DNN)-were developed and optimized using 100 MR images obtained in healthy individuals and validated in a separate dataset of 20 patients with generalized epilepsy, which is an indication for CM-DBS. Models were trained to predict the stereotactic coordinates of the CM using input features, which were x, y, and z coordinates of readily identifiable points from T1-weighted MRI.
RESULTS: The DNN model demonstrated the highest accuracy in predicting CM coordinates, with a mean Euclidean error of 0.88 ± 0.41 mm in the healthy dataset, and 1.12 ± 0.44 mm in the epilepsy dataset. The LR, SVR, and KNN models all performed similarly, although with higher error rates.
CONCLUSIONS: These findings indicate that ML models, particularly DNNs, can accurately predict CM coordinates using standard T1-weighted MRI. This approach reduces dependency on advanced imaging techniques, making CM-DBS more accessible.
METHODS: Four ML models-linear regression (LR), k-nearest neighbor (KNN), support vector regression (SVR), and deep neural network (DNN)-were developed and optimized using 100 MR images obtained in healthy individuals and validated in a separate dataset of 20 patients with generalized epilepsy, which is an indication for CM-DBS. Models were trained to predict the stereotactic coordinates of the CM using input features, which were x, y, and z coordinates of readily identifiable points from T1-weighted MRI.
RESULTS: The DNN model demonstrated the highest accuracy in predicting CM coordinates, with a mean Euclidean error of 0.88 ± 0.41 mm in the healthy dataset, and 1.12 ± 0.44 mm in the epilepsy dataset. The LR, SVR, and KNN models all performed similarly, although with higher error rates.
CONCLUSIONS: These findings indicate that ML models, particularly DNNs, can accurately predict CM coordinates using standard T1-weighted MRI. This approach reduces dependency on advanced imaging techniques, making CM-DBS more accessible.