Pith. sign in

REVIEW 4 major objections 5 minor 80 references

A Deep RL based Framework for Targeted White Matter Tractography

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read GPT-based sequence models can refine and fuse RL tracking policies for tract-specific brain tractography without ground-truth fibers.

desk verdict Solid engineering thesis with a plausible but statistically unsupported claim of improvement over base RL policies; worth refereeing with major revisions. read the letter →

arxiv 2608.12960 v1 pith:LI4FQ5BB submitted 2026-08-13 cs.CV

Ankita Joshi This is my paper · ORCID
classification cs.CV
keywords whitemattertractographyreinforcementlearningpolicyrefinementfusionGPTsequencemodelingtract-specificdiffusionMRInoground-truthtraining
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Tractography reconstructs white-matter pathways from diffusion MRI, but whole-brain tracking produces many false positives and needs a separate segmentation step, while supervised training requires ground-truth fibers that are rarely available. This thesis claims that a tract-specific hybrid of reinforcement learning and supervised sequence modeling can address both problems at once. In the proposed frameworks, a GPT-style decoder-only transformer is trained offline on rollouts from RL tracking agents, represented as state, action, and return-to-go tuples, first on a mixed set of tracts and then fine-tuned per tract, without ever seeing ground-truth streamlines. The refined policy, called Tract-RLFormer, and the fused policy, called TractRLFusion, combining TD3, SAC, and DDPG, are reported to improve Dice and overlap while reducing overreach compared with base policies and whole-brain methods, and to generalize from TractoInferno to HCP and ISMRM data. If the claim is right, tract-specific tractography no longer needs labeled fibers or an explicit segmentation stage, only atlas-derived masks and RL experience.

What carries the argument

The central object is the trajectory token sequence: each timestep contributes a return-to-go scalar, a 334-dimensional state built from spherical-harmonic coefficients of a voxel and its neighbors, mask values, and the last four tracking directions, plus a 3-dimensional action, all processed by a causal decoder-only transformer with a 40-token context. This is the mechanism by which offline RL experience becomes a tract-specific policy: the transformer is pre-trained on mixed-tract trajectories and fine-tuned per tract, and at inference the return-to-go is fixed to a high expert value of 300 so the model generates actions conditioned on the promised return. Around this core, the Mask Refinement Module prunes dilated atlas masks into subject-specific tracking regions, Episodic Data Selection curates trajectories by MDF distance and Q-value, and Multi-Critic Policy Fine-Tuning anchors the fused actor with the original policies' critics.

What would settle it

Train the framework under two conditions: once with the true atlas reference streamlines and once with the same atlas rotated, warped, or replaced by a different bundle atlas, keeping all RL rollouts and training hyperparameters fixed. If Dice, overlap, and overreach on held-out subjects do not degrade substantially in the corrupted-atlas condition, the no-ground-truth claim stands; if they do, the reported gains are artifacts of atlas priors rather than evidence of learned policy refinement.

Watch

Extended reading notes

Core claim

The central claim is that a trajectory-level sequence model can refine and fuse RL tracking policies more effectively than the policies can perform on their own. In Tract-RLFormer, a TD3 agent is trained per tract inside masks produced by the Mask Refinement Module; its rollouts are converted into return-to-go, state, and action trajectories; three decoder layers are pre-trained on a mixed-tract dataset of 150,000 trajectories, and a fourth layer is fine-tuned per tract with a five-step cosine angular loss. TractRLFusion repeats this idea across three policies: Episodic Data Selection keeps trajectories whose streamlines are within a 5 mm mean direct-flip distance of atlas reference streamlines and, across policies, selects the trajectories with the highest predicted Q-value; a GPT-based FusionNet is trained on the curated data and then refined by Multi-Critic Policy Fine-Tuning, in which the TD3, SAC, and DDPG critics add Q-value gradients to the five-step loss. The paper reports the highest Dice among the compared methods on nearly every tract and dataset, with lower overreach than the exploratory SAC policy and higher overlap than conservative TD3 and DDPG, and it attributes the gain to the balance the fused policy strikes between overlap and overreach.

Load-bearing premise

The load-bearing premise is that atlas-derived reference streamlines are accurate enough, after registration, to stand in for ground truth when generating masks, selecting training trajectories, and cleaning tracts; if they are wrong for a subject or tract, the claimed gains could come from the atlas prior rather than from learned policy improvement.

Editorial extensions

If this is right

  • Tract-specific RL policies can be improved without ground-truth fibers, using only atlas-derived masks, RL rollouts, and a return-to-go conditioned sequence model.
  • The same pre-trained GPT backbone, fine-tuned per tract, transfers from TractoInferno training data to HCP and ISMRM test data, so a single foundation model could cover many bundles.
  • Fusing a conservative policy such as TD3 or DDPG with an exploratory policy such as SAC produces a better overlap–overreach balance than any single policy or decision-level voting or averaging.
  • Tract-specific masks alone raise the Dice of even untrained classical DET and PROB trackers, so mask quality is a major lever on tractography accuracy.
  • Because training is offline on rollouts, the framework can be extended to new RL policies or new datasets by re-running trajectory collection and fine-tuning, without re-training from scratch.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the no-ground-truth claim holds, the same offline trajectory-refinement recipe should transfer to other RL-based curve-tracing tasks in medical imaging, such as vessel or airway delineation, where annotated centerlines are scarce.
  • The method's dependence on atlas reference streamlines means its no-ground-truth claim is really no subject-specific ground truth; a testable extension would be to measure how much of the reported Dice gain disappears when the atlas is swapped or misaligned.
  • Because FusionNet outperforms TractSeg even inside TractSeg's own masks, the learned fused policy rather than the superior mask may be doing part of the work; this could be isolated by running the base policies inside the same masks.
  • The return-to-go conditioning, fixed at 300 during inference, is a plausible control knob for the overlap–overreach trade-off; the paper does not explore it, but varying this value could offer a simple user-facing sensitivity dial.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The thesis proposes two tract-specific white-matter tractography frameworks built on a GPT decoder-only sequence model. Tract-RLFormer first trains a TD3 policy inside tract-specific masks generated by a Mask Refinement Module, collects rollouts as (return-to-go, state, action) trajectories, and trains a GPT model with mixed-tract pretraining followed by tract-specific finetuning to obtain a refined tracking policy. TractRLFusion extends this idea to fuse three policies (TD3, SAC, DDPG), using Episodic Data Selection to retain anatomically plausible high-Q trajectories, a FusionNet trained on those selected rollouts, and Multi-Critic Policy Fine-Tuning to refine the fused policy. The paper claims improvements in Dice, overlap, and overreach relative to the base and ensemble policies, and reports generalization across TractoInferno, HCP, and ISMRM datasets, all under a stated premise of not relying on ground-truth fibers for training.

Significance. If the reported improvements were statistically supported, the framework would be a useful contribution: it shows a way to reuse RL rollouts and anatomical atlases to obtain tract-specific tracking policies, and to combine complementary RL policies without a supervised fiber dataset for the final policy. The final evaluation is against external reference tracts, so the headline result is not definitionally circular. The paper also has concrete strengths: extensive validation on public benchmarks, comparison with classical, supervised, and RL baselines, and ablations of MRM, the two-stage training, EDS, and MCPFT. The central weakness is quantitative: most reported differences are small relative to the shown variability, no significance testing is provided, and overreach is not consistently reduced. The no-ground-truth claim also needs qualification because atlas-derived reference streamlines are used for mask supervision, trajectory selection, and final cleaning.

major comments (4)
  1. [§4.3.1, Tables 4.3–4.4] The claim that FusionNet consistently improves Dice and better balances overlap versus overreach is not established by the reported numbers. Several headline differences are within one standard deviation: HCP AF Dice is 55.5±4.8 for FusionNet versus 54.6±4.7 for SAC; HCP CST Dice is 69.4±1.3 versus 69.2±0.8 for πavg; TtoI AF Dice is 53.2±8.9 versus 52.3±8.9 for SAC. Overreach is also often higher for FusionNet than for the least-overreaching baseline: HCP AF OR is 12.3 versus 10.3 for DET, HCP CC OR is 44.8 versus 37.3 for DDPG, TtoI AF OR is 39.1 versus 37.6 for TD3, and TtoI CG OR is 41.6 versus 30.4 for DDPG. No p-values, confidence intervals, or paired tests are reported anywhere, and the Chapter 3 tables report no variability at all. Since the abstract and Section 1.5 state improvement as the contribution, the current evidence is insufficient to support the central claim.
  2. [Abstract; §1.5; §3.3.1.2; §4.2.3; §3.3.4] The claim of training 'without ground-truth fibers' and 'without ground-truth annotations' is misleading as stated. The Mask Refinement Module is trained with binary cross-entropy against a voxel-wise ground truth derived from RecobundlesX atlas reference streamlines, with the text explicitly calling this 'the ground truth for each voxel'. Episodic Data Selection filters trajectories by MDF distance to atlas reference streamlines, and the final tract cleaning uses Fast Streamline Search against atlas reference tracts. The accurate statement is that the method does not use subject-specific ground-truth fiber sets, not that it avoids reference/annotation data entirely. Because the same atlas prior appears in mask generation, trajectory selection, and cleaning, it needs a sensitivity analysis or an ablation that removes FSS/atlas-based selection to demonstrate that the reported gains are not largely attributable to the atlas prior rather than to learned policy refinement.
  3. [§3.4.3, Tables 3.3–3.4] Tract-RLFormer is described as consistently outperforming the TD3 policy from which it was trained, but the reported gaps are very small and no standard deviations or tests are given. Examples include HCP left CG Dice 53.3 versus 53.0 for TD3, HCP right CG Dice 45.6 versus 45.2, and TtoI AF Dice 52.7 versus 51.8. Overreach is not consistently reduced: T-RLF has higher OR than TD3 for TtoI CG left (28.6 versus 27.3), AF right (49.8 versus 46.9), PYT left (17.2 versus 15.9), and CC (32.6 versus 26.1). In several rows classical DET/PROB also exceed T-RLF. These results do not support the Section 3.5 summary statement that the framework improves performance and reduces false positives.
  4. [§4.2.3 and §4.2.5] The fusion mechanism assumes that Q-values from TD3, SAC, and DDPG are comparable across policies, since EDS selects the policy with the maximum expected Q-value and MCPFT aggregates the three critics through Eq. (4.2.3). The three policies were trained with different discount factors, learning rates, and exploration/entropy settings (Table 4.1), so their critic outputs are not calibrated to a common scale. The paper does not discuss this cross-policy comparability issue, which is load-bearing for both the selection and the combined actor loss. A concise experimental justification, for example a study of Q-value distributions per policy or a normalized variant of the selection criterion, is needed.
minor comments (5)
  1. [§3.4, §3.4.2] The text says seven principal tracts are used, but Section 3.4 then adds the Optical Radius tract as an eighth tract; the abbreviation list also defines OR as 'Optical Radius', whereas the standard term is optic radiation.
  2. [Table 3.3] There is a typo in the HCP DET overreach entry ('21..5'), and several other numbers use a non-standard notation for standard deviations in Chapter 4 tables; these should be cleaned up.
  3. [§3.4.4, Table 3.5] The caption refers to 'Section 3.2.1.2', which does not exist; this cross-reference should be corrected.
  4. [§4.3.1] The note explaining why PROB is omitted from some TractoInferno rows is informative, but it should be moved into the table caption or stated before the first table that uses this exclusion.
  5. [Figure 3.5] The caption contains the misspelling 'Tract-RLForemer' and should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claims are evaluated against external ground-truth Dice/overlap metrics, and no prediction in the pipeline reduces by construction to its training inputs.

full rationale

The paper's headline results are not definitionally circular. Final evaluation metrics (Dice, Overlap, Overreach) are computed voxel-wise against external ground-truth tracts using Eqs. 3.4.3-3.4.5, not against the masks, rewards, Q-values, or trajectory-selection criteria used in training. The RL policies are trained with a reward based on fODF alignment and streamline continuity (Eq. 3.3.1/4.2.1), the T-RLF and FusionNet models are trained on RL rollouts with a cosine-distance action loss, and the EDS module selects trajectories using atlas MDF constraints and base-policy Q-values; all of these are inputs to the method rather than the evaluation quantities being claimed. The MRM uses atlas-derived masks as supervision, and FSS cleaning uses atlas reference tracts, but the reported Dice/overlap numbers are measured against held-out ground truth from TractoInferno, HCP, and ISMRM, so the claimed improvements are not forced by construction. Self-citations such as [74] are pointers to the thesis's own preceding chapter (Section 3.3.1.2) and do not supply an unverified load-bearing premise; no uniqueness theorem or ansatz is imported from the authors' prior work. Concerns about statistical significance and dependence on atlas priors are correctness or validation risks, not circularity. The derivation chain is therefore self-contained with respect to the evaluation protocol.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests mostly on empirical engineering choices rather than new physical assumptions. No new particles, forces, or conserved quantities are introduced. The main load-bearing assumptions are the atlas-based reference streamlines used in MRM, EDS, and FSS cleaning, plus the assumption that behavior cloning on RL rollouts with a 5-step action loss improves tracking. These are domain and ad hoc assumptions, not derived guarantees. The free parameters are all tuning decisions; the most consequential are the return-to-go of 300 and the dataset-specific step sizes, because they are tuned on the same Dice and overlap metrics used for the final claims.

free parameters (6)
  • Initial return-to-go R0 for inference = 300
    Chosen among 100, 200, 300, 500, and 600 to maximize Dice in Section 3.3.3.3; directly controls action generation in T-RLF and FusionNet.
  • Dataset-specific tracking step size = 0.375 mm (TractoInferno), 0.468 mm (HCP), 0.75 mm (ISMRM)
    Empirically selected in Section 3.3.4; changes streamline sampling and all downstream metrics.
  • MRM voxel retention threshold = 0.5
    Voxel is kept when predicted probability is greater than 0.5; this is a hand-chosen binary decision threshold for tract masks in Section 3.3.1.2.
  • Minimum trajectory length = 47 transitions (20 mm)
    Used to filter training trajectories and tract output in Sections 3.3.3.1 and 4.2.3; chosen to remove anatomically implausible short streamlines.
  • MDF selection threshold and reference count in EDS = 5 mm, 15 streamlines
    Within-policy selection keeps trajectories with MDF distance less than 5 mm to 15 farthest-streamline references in Section 4.2.3; these shape priors are hand-set.
  • T-RLF architecture context length, heads, and embedding = K=40, n_heads=1, d=128
    Selected by the Table 3.5 ablation on subject 1006 to maximize Dice; affects temporal receptive field and model capacity.
assumptions (5)
  • domain assumption The MDP formulation of streamline tracking with state, action, reward, and termination criteria is a valid model for tractography.
    Adopted from Track-to-Learn in Section 3.3.1.1; the reward in Eq. (3.3.1) is assumed to correlate with anatomical correctness.
  • domain assumption Atlas reference streamlines from RecobundlesX are reliable anatomical priors for mask generation, trajectory selection, and cleaning.
    Used in MRM in Section 3.3.1.2, EDS in Section 4.2.3, and FSS in Section 3.3.4; this is the load-bearing external ground truth behind the no-ground-truth claim.
  • ad hoc to paper A 5-step cosine loss on predicted actions is a good proxy for policy quality.
    Eq. (3.3.2) and Eq. (4.2.2) train the GPT to imitate actions rather than to directly maximize any return estimate; no theorem connects this loss to Dice improvement.
  • ad hoc to paper Behavior cloning on selected rollouts of a base RL policy can yield a policy that outperforms the base policy.
    T-RLF is trained on TD3 rollouts and FusionNet on SAC, TD3, and DDPG rollouts; the paper provides empirical evidence but no formal justification for why distillation improves on the source policy.
  • domain assumption CSD-estimated fODF with 45 spherical harmonic coefficients and up to five peaks captures the fiber geometry needed for tractography.
    This representation is the fixed input to the RL environment and masks in Sections 3.2 and 3.3.1.1; errors in fODF estimation propagate into all downstream claims.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Deep RL based Framework for Targeted White Matter Tractography." pith.science (2026). https://pith.science/paper/LI4FQ5BB

@misc{pith2026260812960,
  author       = {Pith},
  title        = {Pith review of: A Deep RL based Framework for Targeted White Matter Tractography},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LI4FQ5BB}},
  note         = {Machine review of arXiv:2608.12960}
}
read the original abstract

Fiber tractography's ability to reconstruct the brain's structural pathways, has made it a crucial component of modern neuroimaging, enabling detailed, non-invasive mapping of structural connectivity and supporting a wide range of neurological research and clinical applications. However, despite its importance, tractography remains a challenging task due to the inherent complexity of white matter structure and its susceptibility to false positives, which can lead to the misrepresentation of critical pathways. To overcome these limitations, in this thesis, we propose a hybrid framework that integrates reinforcement learning with supervised learning for refining RL policies, specifically tailored for tract-specific tractography. Notably, our framework does not rely on ground-truth fibers for training. Moreover, the tract-specific formulation bypasses the need for an explicit segmentation process, simplifying the overall pipeline. Our work includes two main contributions, each building upon the previous. First, we introduce a hybrid approach that combines reinforcement learning with supervised learning (specifically, GPT-based policy learning) to refine policies in a tract-specific context. Second, we propose a scalable framework for data-driven multi-policy fusion, which leverages the complementary strengths of multiple RL policies to improve tractography performance and robustness. We demonstrate the effectiveness of our framework through extensive validation on benchmark public datasets including TractoInferno, HCP, and ISMRM-2015, highlighting its ability to generalize across data sources and accurately reconstruct brain white matter tracts. We believe that these contributions represent significant advancements in the field of tractography, improving robustness, reliability, and accuracy while reducing dependence on ground-truth annotations.

Figures

Figures reproduced from arXiv: 2608.12960 by the authors.

Figure 1.1
Figure 1.1. Siemens Magnetom Verio 3T MRI Scanner [5]. Different imaging techniques, such as Magnetic Resonance Imaging (MRI) [6], Computed Tomography (CT) [7], and Positron Emission Tomography (PET) [8], provide complementary information about brain health, structure, and function. Among these, MRI is currently the most common imaging modality for brain tissue evaluation because of its excellent soft-tissue contrast and high s… view at source ↗
Figure 1.2
Figure 1.2. Overview of Structural MRI, Functional MRI (fMRI), and Diffusion MRI (dMRI) [PITH_FULL_IMAGE:figures/full_fig_p018_1_2.png] view at source ↗
Figure 1.3
Figure 1.3. Illustration of (i) free isotropic movement in like CSF. (ii) depicts Anisotropic along [PITH_FULL_IMAGE:figures/full_fig_p019_1_3.png] view at source ↗
Figures from the paper (17 more)
Figure 1.4
Figure 1.4. Figure 1.4: Major White Matter tracts [27]. connectivity [PITH_FULL_IMAGE:figures/full_fig_p023_1_4.png]
Figure 1.5
Figure 1.5. Figure 1.5: Tractography process Tractography can be understood as tracking in 3D space (brain volume), constrained by a tracking mask and anatomy. Seed points are placed within this mask, and at each step the algorithm predicts the next direction based on local diffusion infor￾…
Figure 1.6
Figure 1.6. Figure 1.6: Whole-brain tractogram and its segmentation into anatomically meaningful white [PITH_FULL_IMAGE:figures/full_fig_p024_1_6.png]
Figure 1.7
Figure 1.7. Figure 1.7: Illustration of common complex fiber configurations, including crossing, branching, and [PITH_FULL_IMAGE:figures/full_fig_p026_1_7.png]
Figure 2.1
Figure 2.1. Figure 2.1: Illustration of a typical Reinforcement Learning (RL) setup: the agent interacts with [PITH_FULL_IMAGE:figures/full_fig_p034_2_1.png]
Figure 2.2
Figure 2.2. Figure 2.2: Model-free RL algorithms (non-exhaustive) [PITH_FULL_IMAGE:figures/full_fig_p038_2_2.png]
Figure 2.3
Figure 2.3. Figure 2.3: Illustration of actor-critic framework in RL [PITH_FULL_IMAGE:figures/full_fig_p041_2_3.png]
Figure 2.4
Figure 2.4. Figure 2.4: Illustration of Transformer model architecture [ [PITH_FULL_IMAGE:figures/full_fig_p045_2_4.png]
Figure 3.1
Figure 3.1. Figure 3.1: Overview of the proposed Tract-RLFormer framework. (a) An RL agent ( [PITH_FULL_IMAGE:figures/full_fig_p054_3_1.png]
Figure 3.2
Figure 3.2. Figure 3.2: Mask Refinement Module for generating tract-specific masks. (1) The atlas reference [PITH_FULL_IMAGE:figures/full_fig_p056_3_2.png]
Figure 3.3
Figure 3.3. Figure 3.3: Data Representation for T-RLF: Tract specific policy refinement using a trajectory [PITH_FULL_IMAGE:figures/full_fig_p059_3_3.png]
Figure 3.4
Figure 3.4. Figure 3.4: T-RLF training: Visual representation of training Tract-RLFormer for action prediction [PITH_FULL_IMAGE:figures/full_fig_p060_3_4.png]
Figure 3.5
Figure 3.5. Figure 3.5: Visual comparison of reconstructed tracts illustrating (a): Intra-dataset variability, [PITH_FULL_IMAGE:figures/full_fig_p067_3_5.png]
Figure 4.1
Figure 4.1. Figure 4.1: TractRLFusion: A generative framework where anatomically relevant trajectories with high expected returns from various RL agents(πθi ) are combined into a experience dataset (T) to train a generative model. The model leverages critic networks (C(πθi )) to align strea…
Figure 4.2
Figure 4.2. Figure 4.2: Visual depiction of Episodic Data Selection. Row 1 illustrates within-policy trajectory selection using the MDF [75] constraint (Eq. 2) with atlas reference streamlines. Row 2 depicts across-policy selection via the Q-value constraint (Eq. 3), where expected returns …
Figure 4.3
Figure 4.3. Figure 4.3: Visualization of MCPFT module: It further finetunes the FusionNet model using [PITH_FULL_IMAGE:figures/full_fig_p078_4_3.png]
Figure 4.4
Figure 4.4. Figure 4.4: Visualization of (a) right corticospinal tract, and (b) part of the left cingulum tract, [PITH_FULL_IMAGE:figures/full_fig_p081_4_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 71 canonical work pages

  1. [1]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,”Advances in neural information processing systems, vol. 25, 2012

    2012

  2. [2]

    Faster r-cnn: Towards real-time ob- ject detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time ob- ject detection with region proposal networks,”Advances in neural information processing systems, vol. 28, 2015

    work page 2015

  3. [3]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inInternational Conference on Medical image computing and computer-assisted intervention. Springer, 2015, pp. 234–241

    2015

  4. [4]

    Photo-realistic single image super- resolution using a generative adversarial network,

    C. Ledig, L. Theis, F. Huszár, J. Caballero, A. Cunningham, A. Acosta, A. Aitken, A. Tejani, J. Totz, Z. Wanget al., “Photo-realistic single image super- resolution using a generative adversarial network,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 4681–4690

    work page 2017

  5. [5]

    Siemens 3t mri scanners compared,

    I. Block Imaging, “Siemens 3t mri scanners compared,” https://www. blockimaging.com/blog/siemens-3t-mri-scanners-compared, accessed: 2025-12- 07

    work page 2025

  6. [6]

    R. H. Hashemi, W. G. Bradley, and C. J. Lisanti,MRI: the basics. Lippincott Williams & Wilkins, 2010

    work page 2010

  7. [7]

    Computed tomography: principles, design, artifacts, and recent ad- vances,

    J. Hsieh, “Computed tomography: principles, design, artifacts, and recent ad- vances,” 2003. 77 Bibliography 78

    work page 2003

  8. [8]

    Molecular imaging and its biological applications,

    M. E. Phelps, “Molecular imaging and its biological applications,”Eur J Nucl Med Mol Imaging, vol. 31, p. 1544, 2004

    work page 2004

Show all 80 references
  1. [9]

    Multimodality assessment of brain tumors and tumor recurrence,

    W.-D. Heiss, P. Raab, and H. Lanfermann, “Multimodality assessment of brain tumors and tumor recurrence,”Journal of Nuclear Medicine, vol. 52, no. 10, pp. 1585–1600, 2011

    2011

  2. [10]

    A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises,

    S. K. Zhou, H. Greenspan, C. Davatzikos, J. S. Duncan, B. Van Ginneken, A. Madabhushi, J. L. Prince, D. Rueckert, and R. M. Summers, “A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises,”Pr...

    2021

  3. [11]

    The intelligent imaging revolution: artificial intelligence in mri and mrs acquisition and re- construction,

    T. Küstner, C. Qin, C. Sun, L. Ning, and C. M. Scannell, “The intelligent imaging revolution: artificial intelligence in mri and mrs acquisition and re- construction,”Magnetic Resonance Materials in Physics, Biology and Medicine, vol. 37, no. 3, pp. 329–333, 2024

    2024

  4. [12]

    Advances and challenges in fmri and dti techniques,

    R. R. Jha, G. Jaswal, A. Nigam, and A. Bhavsar, “Advances and challenges in fmri and dti techniques,”Intelligent data security solutions for e-health applica- tions, pp. 77–90, 2020

    2020

  5. [13]

    Mr diffusion tensor spectroscopy and imaging,

    P. J. Basser, J. Mattiello, and D. LeBihan, “Mr diffusion tensor spectroscopy and imaging,”Biophysical journal, vol. 66, no. 1, pp. 259–267, 1994

    1994

  6. [14]

    Robust determination of the fibre orientation distribution in diffusion mri: non-negativity constrained super- resolved spherical deconvolution,

    J.-D. Tournier, F. Calamante, and A. Connelly, “Robust determination of the fibre orientation distribution in diffusion mri: non-negativity constrained super- resolved spherical deconvolution,”Neuroimage, vol. 35, no. 4, pp. 1459–1472, 2007

    2007

  7. [15]

    Q-ball imaging,

    D. S. Tuch, “Q-ball imaging,”Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine, vol.52, no. 6, pp. 1358–1372, 2004

    2004

  8. [16]

    Regularized, fast, and robust analytical q-ball imaging,

    M. Descoteaux, E. Angelino, S. Fitzgibbons, and R. Deriche, “Regularized, fast, and robust analytical q-ball imaging,”Magnetic Resonance in Medicine: An Of- Bibliography 79 ficial Journal of the International Society for Magnetic Resonance in Medicine, vol. 58, no. 3, pp. 497–...

    2007

  9. [17]

    An integrated approach to correc- tion for off-resonance effects and subject movement in diffusion mr imaging,

    J. L. Andersson and S. N. Sotiropoulos, “An integrated approach to correc- tion for off-resonance effects and subject movement in diffusion mr imaging,” Neuroimage, vol. 125, pp. 1063–1078, 2016

    2016

  10. [18]

    N4itk: improved n3 bias correction,

    N. J. Tustison, B. B. Avants, P. A. Cook, Y. Zheng, A. Egan, P. A. Yushkevich, and J. C. Gee, “N4itk: improved n3 bias correction,”IEEE transactions on medical imaging, vol. 29, no. 6, pp. 1310–1320, 2010

    2010

  11. [19]

    Mrtrix3: A fast, flexi- bleandopensoftwareframeworkformedicalimageprocessingandvisualisation,

    J.-D. Tournier, R. Smith, D. Raffelt, R. Tabbara, T. Dhollander, M. Pietsch, D. Christiaens, B. Jeurissen, C.-H. Yeh, and A. Connelly, “Mrtrix3: A fast, flexi- bleandopensoftwareframeworkformedicalimageprocessingandvisualisation,” Neuroimage, vol. 202, p. 116137, 2019

    2019

  12. [20]

    The design of sim- pleitk,

    B. C. Lowekamp, D. T. Chen, L. Ibáñez, and D. Blezek, “The design of sim- pleitk,”Frontiers in neuroinformatics, vol. 7, p. 45, 2013

    2013

  13. [21]

    Advanced normalization tools,

    B. B. Avantset al., “Advanced normalization tools,”Insight j, vol. 2, no. 365, pp. 1–35, 2009

    2009

  14. [22]

    Freesurfer,

    B. Fischl, “Freesurfer,”Neuroimage, vol. 62, no. 2, pp. 774–781, 2012

    2012

  15. [23]

    Automated quality control for within and between studies diffusion mri data using a non-parametric framework for movement and distortion correction,

    M. Bastiani, M. Cottaar, S. P. Fitzgibbon, S. Suri, F. Alfaro-Almagro, S. N. Sotiropoulos, S. Jbabdi, and J. L. Andersson, “Automated quality control for within and between studies diffusion mri data using a non-parametric framework for movement and distortion correction,”Neur...

    2019

  16. [24]

    Dtiprep: quality control of diffusion-weighted images,

    I. Oguz, M. Farzinfar, J. Matsui, F. Budin, Z. Liu, G. Gerig, H. J. Johnson, and M. Styner, “Dtiprep: quality control of diffusion-weighted images,”Frontiers in neuroinformatics, vol. 8, p. 4, 2014

    2014

  17. [25]

    Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor mri,

    P. J. Basser and C. Pierpaoli, “Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor mri,”Journal of magnetic res- onance, vol. 213, no. 2, pp. 560–570, 2011. Bibliography 80

    2011

  18. [26]

    Toward a quantitative assessment of diffusion anisotropy,

    C. Pierpaoli and P. J. Basser, “Toward a quantitative assessment of diffusion anisotropy,”Magnetic resonance in Medicine, vol. 36, no. 6, pp. 893–906, 1996

    1996

  19. [27]

    Inference of a human brain fiber bundle atlas from high angular resolution diffusion imaging,

    P. B. G. Alvez, “Inference of a human brain fiber bundle atlas from high angular resolution diffusion imaging,” Ph.D. dissertation, Université Paris Sud-Paris XI, 2011

    2011

  20. [28]

    Tractseg-fast and accurate white matter tract segmentation,

    J. Wasserthal, P. Neher, and K. H. Maier-Hein, “Tractseg-fast and accurate white matter tract segmentation,”NeuroImage, vol. 183, pp. 239–253, 2018

    2018

  21. [29]

    The challenge of mapping the human connectome based on diffusion tractography,

    K. H. Maier-Heinet al., “The challenge of mapping the human connectome based on diffusion tractography,”Nature communications, vol. 8, no. 1, p. 1349, 2017

    2017

  22. [30]

    Diffusion mri fiber tractography of the brain,

    B. Jeurissen, M. Descoteaux, S. Mori, and A. Leemans, “Diffusion mri fiber tractography of the brain,”NMR in Biomedicine, vol. 32, no. 4, p. e3785, 2019

    2019

  23. [31]

    Challenges in diffusion mri tractography–lessons learned from international benchmark competitions,

    K. G. Schilling, A. Daducci, K. Maier-Hein, C. Poupon, J.-C. Houde, V. Nath, A. W. Anderson, B. A. Landman, and M. Descoteaux, “Challenges in diffusion mri tractography–lessons learned from international benchmark competitions,” Magnetic resonance imaging, vol. 57, pp. 194–209, 2019

    2019

  24. [32]

    Tractography and machine learning: Current state and open challenges,

    P. Poulin, D. Jörgens, P.-M. Jodoin, and M. Descoteaux, “Tractography and machine learning: Current state and open challenges,”Magnetic resonance imag- ing, vol. 64, pp. 37–48, 2019

    2019

  25. [33]

    Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging,

    S. Mori, B. J. Crain, V. P. Chacko, and P. C. Van Zijl, “Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging,” Annals of Neurology: Official Journal of the American Neurological Association and the Child Neurology Society, vol. 45, no. 2...

    1999

  26. [34]

    In vivo fiber tractography using dt-mri data,

    P. J. Basser, S. Pajevic, C. Pierpaoli, J. Duda, and A. Aldroubi, “In vivo fiber tractography using dt-mri data,”Magnetic resonance in medicine, vol. 44, no. 4, pp. 625–632, 2000

    2000

  27. [35]

    Characterization and Bibliography 81 propagation of uncertainty in diffusion-weighted mr imaging,

    T. E. Behrens, M. W. Woolrich, M. Jenkinson, H. Johansen-Berg, R. G. Nunes, S. Clare, P. M. Matthews, J. M. Brady, and S. M. Smith, “Characterization and Bibliography 81 propagation of uncertainty in diffusion-weighted mr imaging,”Magnetic Reso- nance in Medicine: An Official ...

    2003

  28. [36]

    Improved probabilistic streamlines tractography by 2nd order integration over fibre orientation distri- butions,

    J. D. Tournier, F. Calamante, A. Connellyet al., “Improved probabilistic streamlines tractography by 2nd order integration over fibre orientation distri- butions,” inProceedings of the international society for magnetic resonance in medicine, vol. 1670. Stockholm, 2010, p. 2010

    2010

  29. [37]

    A framework for a streamline-based probabilistic index of connectivity (pico) using a structural interpretation of mri diffusion measurements,

    G. J. Parker, H. A. Haroon, and C. A. Wheeler-Kingshott, “A framework for a streamline-based probabilistic index of connectivity (pico) using a structural interpretation of mri diffusion measurements,”Journal of Magnetic Resonance Imaging: An Official Journal of the Internatio...

    2003

  30. [38]

    Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging,

    T. E. Behrens, H. Johansen-Berg, M. W. Woolrich, S. M. Smith, C. A. Wheeler- Kingshott, P. A. Boulby, G. J. Barker, E. Sillery, K. Sheehan, O. Ciccarelliet al., “Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging,”Nature neuroscience,...

    2003

  31. [39]

    Gibbs tracking: a novel ap- proach for the reconstruction of neuronal pathways,

    B. W. Kreher, I. Mader, and V. G. Kiselev, “Gibbs tracking: a novel ap- proach for the reconstruction of neuronal pathways,”Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Reso- nance in Medicine, vol. 60, no. 4, pp. 953–963, 2008

    2008

  32. [40]

    Anatomically- constrained tractography: improved diffusion mri streamlines tractography through effective use of anatomical information,

    R. E. Smith, J.-D. Tournier, F. Calamante, and A. Connelly, “Anatomically- constrained tractography: improved diffusion mri streamlines tractography through effective use of anatomical information,”Neuroimage, vol. 62, no. 3, pp. 1924–1938, 2012

    1924

  33. [41]

    Towards quanti- tative connectivity analysis: reducing tractography biases,

    G. Girard, K. Whittingstall, R. Deriche, and M. Descoteaux, “Towards quanti- tative connectivity analysis: reducing tractography biases,”Neuroimage, vol. 98, pp. 266–278, 2014. Bibliography 82

    2014

  34. [42]

    Tractoinferno-a large-scale, open-source, multi-site database for machine learning dmri tractography,

    P. Poulinet al., “Tractoinferno-a large-scale, open-source, multi-site database for machine learning dmri tractography,”Scientific Data, vol. 9, no. 1, p. 725, 2022

    2022

  35. [43]

    A machine learning based approach to fiber tractography using classifier voting,

    P. F. Neher, M. Götz, T. Norajitra, C. Weber, and K. H. Maier-Hein, “A machine learning based approach to fiber tractography using classifier voting,” in International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2015, pp. 45–52

    2015

  36. [44]

    Learn to track: deep learning for tractography,

    P. Poulinet al., “Learn to track: deep learning for tractography,” inMICCAI 2017: 20th International Conference, Quebec City, QC, Canada, September 11- 13, 2017, Proceedings, Part I 20. Springer, 2017, pp. 540–547

    2017

  37. [45]

    Deeptract: A probabilistic deep learning frame- work for white matter fiber tractography,

    I. Benou and T. Riklin Raviv, “Deeptract: A probabilistic deep learning frame- work for white matter fiber tractography,” inMICCAI: Shenzhen, China, Octo- ber 13–17, 2019. Springer, 2019, pp. 626–635

    2019

  38. [46]

    Data-driven fiber tractography with neural networks,

    V. Wegmayr, G. Giuliari, S. Holdener, and J. Buhmann, “Data-driven fiber tractography with neural networks,” in2018 IEEE 15th international symposium on biomedical imaging (ISBI 2018). IEEE, 2018, pp. 1030–1033

    2018

  39. [47]

    Entrack: Probabilistic spherical regres- sion with entropy regularization for fiber tractography,

    V. Wegmayr and J. M. Buhmann, “Entrack: Probabilistic spherical regres- sion with entropy regularization for fiber tractography,”International Journal of Computer Vision, vol. 129, no. 3, pp. 656–680, 2021

    2021

  40. [48]

    Track-to-learn: A general framework for tractography with deep reinforcement learning,

    A. Thébergeet al., “Track-to-learn: A general framework for tractography with deep reinforcement learning,”MIA, vol. 72, p. 102093, 2021

    2021

  41. [49]

    What matters in reinforcement learning for tractography,

    A. Théberge, C. Desrosiers, A. Boré, M. Descoteaux, and P.-M. Jodoin, “What matters in reinforcement learning for tractography,”MIA, vol. 93, p. 103085, 2024

    2024

  42. [50]

    Tractoracle: towards an anatomically-informed reward function for rl-based tractography,

    A. Théberge, M. Descoteaux, and P.-M. Jodoin, “Tractoracle: towards an anatomically-informed reward function for rl-based tractography,” inInterna- tional Conference on Medical Image Computing and Computer-Assisted Inter- vention. Springer, 2024, pp. 476–486. Bibliography 83

    2024

  43. [51]

    Motor and language dti fiber tracking combined with intraoperative subcortical mapping for surgical removal of gliomas,

    L. Bello, A. Gambini, A. Castellano, G. Carrabba, F. Acerbi, E. Fava, C. Gius- sani, M. Cadioli, V. Blasi, A. Casarottiet al., “Motor and language dti fiber tracking combined with intraoperative subcortical mapping for surgical removal of gliomas,”Neuroimage, vol. 39, no. 1, p...

    2008

  44. [52]

    Bundle-specific tractography with incorporated anatomical and orientational priors,

    F. Rheault, E. St-Onge, J. Sidhu, K. Maier-Hein, N. Tzourio-Mazoyer, L. Petit, and M. Descoteaux, “Bundle-specific tractography with incorporated anatomical and orientational priors,”Neuroimage, vol. 186, pp. 382–398, 2019

    2019

  45. [53]

    Openai gym,

    G. Brockman, V. Cheung, L. Pettersson, J. Schneider, J. Schulman, J. Tang, and W. Zaremba, “Openai gym,”arXiv preprint arXiv:1606.01540, 2016

    2016 arXiv

  46. [54]

    Policy gradient meth- ods for reinforcement learning with function approximation,

    R. S. Sutton, D. McAllester, S. Singh, and Y. Mansour, “Policy gradient meth- ods for reinforcement learning with function approximation,”Advances in neural information processing systems, vol. 12, 1999

    1999

  47. [55]

    Continuous control with deep reinforcement learning,

    T. Lillicrap, “Continuous control with deep reinforcement learning,”arXiv preprint arXiv:1509.02971, 2015

    2015 arXiv

  48. [56]

    Addressing function approximation error in actor-critic methods,

    S. Fujimoto, H. Hoof, and D. Meger, “Addressing function approximation error in actor-critic methods,” inICML. PMLR, 2018, pp. 1587–1596

    2018

  49. [57]

    Soft actor-critic: Off-policy maximumentropydeepreinforcementlearningwithastochasticactor,

    T. Haarnoja, A. Zhou, P. Abbeel, and S. Levine, “Soft actor-critic: Off-policy maximumentropydeepreinforcementlearningwithastochasticactor,” inICML. PMLR, 2018, pp. 1861–1870

    2018

  50. [58]

    Learning internal repre- sentations by error propagation,

    D. E. Rumelhart, G. E. Hinton, R. J. Williamset al., “Learning internal repre- sentations by error propagation,” 1985

    1985

  51. [59]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural compu- tation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  52. [60]

    Attention is all you need,

    A. Vaswaniet al., “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  53. [61]

    Improving lan- guage understanding by generative pre-training,

    A. Radford, K. Narasimhan, T. Salimans, I. Sutskeveret al., “Improving lan- guage understanding by generative pre-training,” 2018. Bibliography 84

    2018

  54. [62]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” inProceedings of the 2019 conference of the North American chapter of the association for com- putational linguistics: human language technolog...

    2019

  55. [63]

    Informer: Beyond efficient transformer for long sequence time-series forecasting,

    H. Zhou, Zhanget al., “Informer: Beyond efficient transformer for long sequence time-series forecasting,” inProceedings of the AAAI conference on artificial in- telligence, vol. 35, no. 12, 2021, pp. 11106–11115

    2021

  56. [64]

    Highly accurate protein structure prediction with alphafold,

    J. Jumper, R. Evans, A. Pritzel, T. Greenet al., “Highly accurate protein structure prediction with alphafold,”nature, vol. 596, no. 7873, pp. 583–589, 2021

    2021

  57. [65]

    Evaluating large language models trained on code,

    M. Chen, “Evaluating large language models trained on code,”arXiv preprint arXiv:2107.03374, 2021

    2021 arXiv

  58. [66]

    An image is worth 16x16 words: Transformers for image recog- nition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Un- terthineret al., “An image is worth 16x16 words: Transformers for image recog- nition at scale,”arXiv preprint arXiv:2010.11929, 2020

    2010 arXiv

  59. [67]

    The human connectome project: a data acquisition perspective,

    D.C.VanEssen, K.Ugurbil, E.Auerbach, D.Barch, T.E.Behrens, R.Bucholz, A. Chang, L. Chen, M. Corbetta, S. W. Curtisset al., “The human connectome project: a data acquisition perspective,”Neuroimage, vol. 62, no. 4, pp. 2222– 2231, 2012

    2012

  60. [68]

    “scilpy,” https://github.com/scilus/scilpy

  61. [69]

    Population average atlas for recobundlesx,

    F. Rheault, “Population average atlas for recobundlesx,” May 2023. [Online]. Available: https://doi.org/10.5281/zenodo.7950602

    2023 doi

  62. [70]

    Decision transformer: Reinforcement learning via sequence modeling,

    L. Chenet al., “Decision transformer: Reinforcement learning via sequence modeling,”NeurIPS, vol. 34, pp. 15084–15097, 2021

    2021

  63. [71]

    Fast streamline search: An exact technique for diffusion mri tractography,

    E. St-Onge, E. Garyfallidis, and D. L. Collins, “Fast streamline search: An exact technique for diffusion mri tractography,”Neuroinformatics, vol. 20, no. 4, pp. 1093–1104, 2022. Bibliography 85

    2022

  64. [72]

    Advancements in diffusion mri tractography for neurosurgery,

    K. Kamagata, C. Andica, W. Uchida, K. Takabayashi, Y. Saito, M. Lukies, A. Hagiwara, S. Fujita, T. Akashi, A. Wadaet al., “Advancements in diffusion mri tractography for neurosurgery,”Investigative Radiology, vol. 59, no. 1, pp. 13–25, 2024

    2024

  65. [73]

    Offline imitation learning from multiple baselines with applications to compiler optimization,

    T. V. Marinov, A. Agarwal, and M. Trofin, “Offline imitation learning from multiple baselines with applications to compiler optimization,”arXiv preprint arXiv:2403.19462, 2024

    2024 arXiv

  66. [74]

    Tract-rlformer: A tract-specific rl policy based decoder-only trans- former network,

    A. Joshi, A. Sharma, A. Goel, R. R. Jha, C. K. Ahuja, A. Bhavsar, and A. Nigam, “Tract-rlformer: A tract-specific rl policy based decoder-only trans- former network,” inInternational Conference on Pattern Recognition. Springer, 2024, pp. 258–275

    2024

  67. [75]

    Quickbundles, a method for tractography simplification,

    E. Garyfallidis, M. Brett, M. M. Correia, G. B. Williams, and I. Nimmo-Smith, “Quickbundles, a method for tractography simplification,”Frontiers in neuro- science, vol. 6, p. 175, 2012

    2012

  68. [76]

    Farthest streamline sampling for the uniform distribution of forearm muscle fiber tracts from diffusion tensor imaging,

    Y. Li, S. Ma, J. Zhao, Q. Li, and X. Sheng, “Farthest streamline sampling for the uniform distribution of forearm muscle fiber tracts from diffusion tensor imaging,”arXiv preprint arXiv:2306.13969, 2023

    2023 arXiv

  69. [77]

    Multi-critic actor learning: Teaching rl policies to act with style,

    S. Mysore, G. Cheng, Y. Zhao, K. Saenko, and M. Wu, “Multi-critic actor learning: Teaching rl policies to act with style,” inInternational Conference on Learning Representations, 2022

    2022

  70. [78]

    Mrtrix: diffusion tractography in crossing fiber regions,

    J.-D. Tournier, F. Calamante, and A. Connelly, “Mrtrix: diffusion tractography in crossing fiber regions,”International journal of imaging systems and technol- ogy, vol. 22, no. 1, pp. 53–66, 2012

    2012

  71. [79]

    Fiesta: Autoen- coders for accurate fiber segmentation in tractography,

    F. Dumais, J. H. Legarreta, C. Lemaire, P. Poulin, F. Rheault, L. Petit, M. Barakovic, S. Magon, M. Descoteaux, P.-M. Jodoinet al., “Fiesta: Autoen- coders for accurate fiber segmentation in tractography,”NeuroImage, vol. 279, p. 120288, 2023. Bibliography 86

    2023

  72. [80]

    On the opportunities and risks of foundation models,

    R. Bommasani, “On the opportunities and risks of foundation models,”arXiv preprint arXiv:2108.07258, 2021

    2021 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.