DeLTA-BIT: an open-source probabilistic tractography-based deep learning framework for thalamic targeting in functional neurological disorders
arxiv(2023)
摘要
In the last years there has been a growing interest in the applications of diffusion tractography for target identification in neurofunctional disorders for an increasingly tailored approach. The growing diffusion of well-established neurosurgical procedures as deep brain stimulation (DBS) and trans-cranial Magnetic Resonance-guided Focused Ultrasound (tcMRgFUS) favored this trend. Tractography can indeed provide more accurate, patient-specific, information about the targeted region if compared to stereotactic atlases. On the other hand, this tractography-based approachs is not very physician-friendly, and its heavily time consuming since needs several hours for Magnetic Resonance Imaging (MRI) data processing. In this study we propose a novel open-source deep learning framework called DeLTA-BIT (acronym of Deep-learning Local TrActography for BraIn Targeting) for reconstructing thalamic probabilistic tractography maps enabling a personalized (patient-specific) target identification for functional neurological disorders. The proposed framework exploits two convolutional neural networks (CNNs) to segment the thalamus and reconstruct the connectivity map between each voxel inside the thalamus and any user-specified target cortical or subcortical masks. These CNNs were trained, validated, and tested on the datasets from the Human Connectome Project (HCP) database. Our results are comparable with the state-of-the-art data-driven segmentation and probabilistic tractography and demonstrate the high performance of proposed networks for personalized thalamic target identification with a Dice coefficient of 0.93 +/- 0.01 for thalamus segmentation and a Dice Similarity Coefficient for the projection to the precentral gyrus from the thalamus of 0.80 +/- 0.06. The code of these models is freely available on GitHub.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要