Virtual dissection of diffusion MRI tractograms is cumbersome and needs extensive knowledge of white matter anatomy. This virtual dissection often requires several inclusion and exclusion ...regions-of-interest that make it a process that is very hard to reproduce across experts. Having automated tools that can extract white matter bundles for tract-based studies of large numbers of people is of great interest for neuroscience and neurosurgical planning. The purpose of our proposed method, named RecoBundles, is to segment white matter bundles and make virtual dissection easier to perform. This can help explore large tractograms from multiple persons directly in their native space. RecoBundles leverages latest state-of-the-art streamline-based registration and clustering to recognize and extract bundles using prior bundle models. RecoBundles uses bundle models as shape priors for detecting similar streamlines and bundles in tractograms. RecoBundles is 100% streamline-based, is efficient to work with millions of streamlines and, most importantly, is robust and adaptive to incomplete data and bundles with missing components. It is also robust to pathological brains with tumors and deformations. We evaluated our results using multiple bundles and showed that RecoBundles is in good agreement with the neuroanatomical experts and generally produced more dense bundles. Across all the different experiments reported in this paper, RecoBundles was able to identify the core parts of the bundles, independently from tractography type (deterministic or probabilistic) or size. Thus, RecoBundles can be a valuable method for exploring tractograms and facilitating tractometry studies.
Display omitted
We present a mathematical construction for the restricted Boltzmann machine (RBM) that does not require specifying the number of hidden units. In fact, the hidden layer size is adaptive and can grow ...during training. This is obtained by first extending the RBM to be sensitive to the ordering of its hidden units. Then, with a carefully chosen definition of the energy function, we show that the limit of infinitely many hidden units is well defined. As with RBM, approximate maximum likelihood training can be performed, resulting in an algorithm that naturally and adaptively adds trained hidden units during learning. We empirically study the behavior of this infinite RBM, showing that its performance is competitive to that of the RBM, while not requiring the tuning of a hidden layer size.
Display omitted
•New online system to evaluate tractography pipelines.•New connectivity-based quantitative measures (valid and invalid bundles/connections and average bundle coverage).•More than ...57,000 different combinations of tractography pipelines compared from the FiberCup dataset.•Take-home messages for tractography users and insights on remaining open questions for the community.
We have developed the Tractometer: an online evaluation and validation system for tractography processing pipelines. One can now evaluate the results of more than 57,000 fiber tracking outputs using different acquisition settings (b-value, averaging), different local estimation techniques (tensor, q-ball, spherical deconvolution) and different tracking parameters (masking, seeding, maximum curvature, step size). At this stage, the system is solely based on a revised FiberCup analysis, but we hope that the community will get involved and provide us with new phantoms, new algorithms, third party libraries and new geometrical metrics, to name a few. We believe that the new connectivity analysis and tractography characteristics proposed can highlight limits of the algorithms and contribute in solving open questions in fiber tracking: from raw data to connectivity analysis. Overall, we show that (i) averaging improves quality of tractography, (ii) sharp angular ODF profiles helps tractography, (iii) seeding and multi-seeding has a large impact on tractography outputs and must be used with care, and (iv) deterministic tractography produces less invalid tracts which leads to better connectivity results than probabilistic tractography.
Fiber tractography using machine learning Neher, Peter F.; Côté, Marc-Alexandre; Houde, Jean-Christophe ...
NeuroImage,
September 2017, 2017-09-00, 20170901, Letnik:
158
Journal Article
Recenzirano
Odprti dostop
We present a fiber tractography approach based on a random forest classification and voting process, guiding each step of the streamline progression by directly processing raw diffusion-weighted ...signal intensities. For comparison to the state-of-the-art, i.e. tractography pipelines that rely on mathematical modeling, we performed a quantitative and qualitative evaluation with multiple phantom and in vivo experiments, including a comparison to the 96 submissions of the ISMRM tractography challenge 2015. The results demonstrate the vast potential of machine learning for fiber tractography.
•First machine learning-driven approach to fiber tractography.•Processing of the raw signal. No mathematical modeling and inverse problem solving.•Extensive evaluation using publicly available in vivo and in silico data.•Highly promising results compared to over 100 tractography pipelines.
Tractography based on non-invasive diffusion imaging is central to the study of human brain connectivity. To date, the approach has not been systematically validated in ground truth studies. Based on ...a simulated human brain data set with ground truth tracts, we organized an open international tractography challenge, which resulted in 96 distinct submissions from 20 research groups. Here, we report the encouraging finding that most state-of-the-art algorithms produce tractograms containing 90% of the ground truth bundles (to at least some extent). However, the same tractograms contain many more invalid than valid bundles, and half of these invalid bundles occur systematically across research groups. Taken together, our results demonstrate and confirm fundamental ambiguities inherent in tract reconstruction based on orientation information alone, which need to be considered when interpreting tractography and connectivity results. Our approach provides a novel framework for estimating reliability of tractography and encourages innovation to address its current limitations.
In this work, we propose a diffusion MRI protocol for mining Parkinson's disease diffusion MRI datasets and recover robust disease-specific biomarkers. Using advanced high angular resolution ...diffusion imaging (HARDI) crossing fiber modeling and tractography robust to partial volume effects, we automatically dissected 50 white matter (WM) fascicles. These fascicles connect deep nuclei (thalamus, putamen, pallidum) to different cortical functional areas (associative, motor, sensorimotor, limbic), basal forebrain and substantia nigra. Then, among these 50 candidate WM fascicles, only the ones that passed a test-retest reproducibility procedure qualified for further tractometry analysis. Leveraging the unique 2-timepoints test-retest Parkinson's Progression Markers Initiative (PPMI) dataset of over 600 subjects, we found statistically significant differences in tract profiles along the subcortico-cortical pathways between Parkinson's disease patients and healthy controls. In particular, significant increases in FA, apparent fiber density, tract-density and generalized FA were detected in some locations of the nigro-subthalamo-putaminal-thalamo-cortical pathway. This connection is one of the major motor circuits balancing the coordination of motor output. Detailed and quantifiable knowledge on WM fascicles in these areas is thus essential to improve the quality and outcome of Deep Brain Stimulation, and to target new WM locations for investigation.
Text-based games offer a challenging test bed to evaluate virtual agents at language understanding, multi-step problem-solving, and common-sense reasoning. However, speed is a major limitation of ...current text-based games, capping at 300 steps per second, mainly due to the use of legacy tooling. In this work we present TextWorldExpress, a high-performance simulator that includes implementations of three common text game benchmarks that increases simulation throughput by approximately three orders of magnitude, reaching over one million steps per second on common desktop hardware. This significantly reduces experiment runtime, enabling billion-step-scale experiments in about one day.
While large language models (LLMs) have been increasingly deployed across tasks in language understanding and interactive decision-making, their impressive performance is largely due to the ...comprehensive and in-depth domain knowledge embedded within them. However, the extent of this knowledge can vary across different domains. Existing methods often assume that LLMs already possess such comprehensive and in-depth knowledge of their environment, overlooking potential gaps in their understanding of actual world dynamics. To address this gap, we introduce Discover, Verify, and Evolve (DiVE), a framework that discovers world dynamics from a small number of demonstrations, verifies the correctness of these dynamics, and evolves new, advanced dynamics tailored to the current situation. Through extensive evaluations, we analyze the impact of each component on performance and compare the automatically generated dynamics from DiVE with human-annotated world dynamics. Our results demonstrate that LLMs guided by DiVE can make better decisions, achieving rewards comparable to human players in the Crafter environment.
We introduce Language Feedback Models (LFMs) that identify desirable behaviour - actions that help achieve tasks specified in the instruction - for imitation learning in instruction following. To ...train LFMs, we obtain feedback from Large Language Models (LLMs) on visual trajectories verbalized to language descriptions. First, by using LFMs to identify desirable behaviour to imitate, we improve in task-completion rate over strong behavioural cloning baselines on three distinct language grounding environments (Touchdown, ScienceWorld, and ALFWorld). Second, LFMs outperform using LLMs as experts to directly predict actions, when controlling for the number of LLM output tokens. Third, LFMs generalize to unseen environments, improving task-completion rate by 3.5-12.0% through one round of adaptation. Finally, LFM can be modified to provide human-interpretable feedback without performance loss, allowing human verification of desirable behaviour for imitation learning.