List of Papers Browse by Subject Areas Author List
Abstract
Tremor is a movement disorder characterized by involuntary, rhythmic oscillations of body parts and is a hallmark of several neurological conditions, including Parkinson’s disease and essential tremor. Elucidating its underlying mechanisms relies heavily on mouse models, which offer genetic manipulability and translational relevance to human neural circuitry. Accordingly, these models are indispensable for studying tremor pathophysiology. So far, electromyography and accelerometers have been used as methods to quantitatively observe tremors in mice.
However, these methods have several drawbacks, such as high costs and complex setups. In particular, the invasive surgical implantation of devices causes significant stress to the animals. Although RGB-based methods offer non-invasive and cost-effective alternatives, they often lack the sensitivity required to detect subtle tremors. Therefore, this paper addresses these challenges by achieving mouse tremor severity estimation using conventional RGB cameras only. To address the challenging task of isolating tremor-related vibrations while the mouse itself is also in motion, our pipeline incorporates segmentation-based preprocessing to extract the mouse region and a Tremor Score Estimation Module that captures subtle tremors with high sensitivity. In the experiments, we assessed tremors in unrestrained mice using a non-invasive method with two standard cameras. The results demonstrated a strong correlation with accelerometer measurements and confirmed that the method accurately captured the intensity-dependent characteristics of tremors. The project page is available at https://isogawalab.github.io/Video-based-Tremor-Analysis-Project/.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/0587_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026/supp/0587_supp.zip
Link to the Code Repository
https://github.com/hotatetabetai/Markerless-Video-based-Tremor-Analysis
Link to the Dataset(s)
N/A
BibTex
@InProceedings{KosYot_Toward_MICCAI2026,
author = { Koshimoto, Yota AND Tsukahara, Akihiro AND Moriwaki, Yasuhiro AND Isogawa, Mariko},
title = { { Toward Markerless Video-Based Tremor Analysis: Objective Quantification of Pathological Tremor in Mouse Preclinical Models } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16896},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The authors propose a framework to estimate tremor in a Parkinson’s Disease mouse model using two standard RGB cameras, DeepLabCut for animal pose estimation, and a segementation module to distinguish between voluntary movement and tremor.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
The suggested framework largely improves current camera-based tremor identification accuracy to be comparable with accelerometer data. The framework is non-invasisive, cost-efficient, and requires only little manual labor.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
The paper does not mention any limitations of their proposed framework, e. g. how does the camera frame rate and lighting affect the pose estimation accuracy and subsequent pipeline. Could additional cameras further improve the accuracy? How does accelerometer data compare to other motion capture data?
The paper does not provide any information on ethics approval for the data collection.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(5) Accept — should be accepted, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper is well-written and clear. The suggested methodology is straight-forward and could have high translational impact.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
The authors addressed my minor concerns in the rebuttal.
Review #2
- Please describe the contribution of the paper
The paper presents a non-invasive framework for quantifying tremor severity in mice using two standard side-view RGB cameras. The pipeline combines segmentation, head pose tracking, confidence-based multi-view selection, a motion artefact suppression module, and prominence-based scoring of vertical head oscillations. The authors also contribute by building a dataset of synchronized RGB and accelerometer data (not shared) and by presenting a study examining how the score correlates with accelerometer-derived tremor measurements.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
-
The paper addresses a practical and important preclinical problem: tremor quantification without invasive sensors, specifically in mice. The problem formulation is also well defined, as the 2 concrete technical challenges are depicted: separating tremor from voluntary locomotion and capturing anti-gravity tremor that may be missed from top-down views. The authors provide implementation details in section 3, but they could be more complete to promote reproducibility.
-
The evaluation is plausible, with minimal proof of concept (as depicted in Table 1). The paper compares against accelerometer PSD, compares with human subjective scoring, includes ablations, and includes a threshold sensitivity analysis. Also, the harmaline dose-response analysis is a meaningful validation, since the method is not only correlated with the sensor signal, but also tracks an expected biological effect, which increases confidence in the score.
-
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
-
The system combines existing ingredients, including SAM 2 for segmentation, DeepLabCut for pose estimation, confidence-based view selection, thresholded movement filtering, and peak prominence scoring. The contribution is useful and well-engineered, but the conceptual advance over prior art feels incremental over Ajima et al. for objective microtremor detection and Ni et al. for motion and tremor tracking.
-
The dataset is relatively limited, comprising 56 sequences from 28 mice collected in a single experimental setup. The manuscript also does not clearly specify whether the thresholds were tuned using held-out animals, making it difficult to assess the risk of overfitting. As a result, the extent to which the method generalizes across sessions, acquisition setups, and posture variations remains unclear. In addition, the subjective evaluation appears to rely on only one experienced evaluator. A stronger analysis of rater consistency would make the comparison with human assessment more convincing.
-
The statistical reporting is somewhat limited. Correlation and one-way ANOVA are useful, but confidence intervals, significance testing between methods, and more detailed error analysis would strengthen the claims.
-
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
-
If possible, correct Fig.1.From the I_front or I_back images, you get the sense that the images are captured horizontally (as stated in section 3.1). The picture, near the accelerometer and the mice, has cameras in different height poses (may mislead);
-
Minor typos or inconsistencies: Introduction section, the phrase “proposes a framework for estimating estimating mouse tremor severity”
-
The remaining suggestions are presented in the previous comments.
-
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper proposes a solution to a meaningful problem and presents a practical system with encouraging validation against accelerometer data. However, for MICCAI, the methodological novelty is modest, the experimental scale is limited, and the evidence of robustness and generalization remains insufficient. I would be most open to changing my score if the rebuttal clearly explains the protocol, makes a stronger case that the baseline comparison is fair, and better argues why this level of novelty, following the MICCAI guidance that reviewers should weigh novelty, soundness, and application impact together rather than focusing on only one axis.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
I’d change my decision to Borderline Accept. After reading the rebuttal, I am satisfied that the authors have addressed my main post-review concerns sufficiently for acceptance. In particular, the additional clarification about thresholds that were selected on a held-out pilot subset (although without further details), the additional confidence intervals and significance tests, and the stronger explanation of the baseline comparison make the experimental evidence more convincing than in the original submission. Some concerns remain: the methodological novelty is still moderate, the dataset is limited to a single acquisition setup, and the human evaluation is based on a single rater. Therefore, I do not view the paper as a very strong methodological contribution. However, the problem is meaningful, the validation against accelerometer PSD is encouraging, and the dose-response analysis supports practical relevance.
Review #3
- Please describe the contribution of the paper
The authors propose a computational pipeline to measure mouse tremor using video data. Specifically, their pipeline includes following phases: recording the mouse with two cameras (front and back views), segmenting the mouse via SAM2, using DeepLabCut to trace head location, filtering out frames with large movements, and counting the number of cases where a local maximum in the y-direction exceeds a local minimum based on a predefined threshold. The experimental results show a better correlation with PSD (ground truth) than human evaluation. The effectiveness is also demonstrated through pharmacological validation by comparing dosage levels with the proposed tremor score.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.The pipeline is straightforward. Filtering out irrelevant data points by thresholding can improve the precision of the downstream task. 2.The manuscript is easy to understand and practically addresses the tremor monitoring problem.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.The authors use wording that may imply that a new module is introduced (e.g., “the proposed method introduces a segmentation module”). However, this segmentation module is essentially a combination of SAM2, DeepLabCut, and thresholding. The authors should consider using different wording in this case.
2.The authors rely on head position movement (x, y) throughout the entire pipeline (for both filtering and tremor detection). However, the depth dimension is completely omitted, even though the mouse moves in 3D space.
3.While hyperparameter studies are provided for the two thresholds, it would be better to provide the corresponding data distributions to justify the selected ranges for the ablation studies.
4.It would be more precise to define the thresholds in physical units (e.g., cm or mm) rather than pixels. Using pixels makes the algorithm difficult to reproduce, since different cameras have different intrinsic parameters.
5.Please justify why PSD can be assumed to accurately capture tremor characteristics and therefore serve as ground truth. It is also unclear why human evaluation shows lower correlation than the proposed method. It would be helpful to discuss cases missed by human evaluation, or whether some cases are false positives from PSD and the proposed method.
6.The baseline setting is somewhat questionable. Ni et al. use an overhead view and track the animal’s center, whereas the authors use a side view while also tracking the animal’s center. Since tracking is performed in different planes, it is unclear whether it is fair to conclude that Ni et al. does not work in this new setting.
7.All module ablation studies show only minimal performance decreases. For example, without the multi-view setting, the correlation is still 0.82.Does this imply that pose estimation still works well in the back view despite self-occlusion? Similarly, without the movement detection module, the correlation remains 0.78.Does this suggest that detecting subtle movement is not particularly necessary? It would also be useful to provide implementation details for each ablation setting.
8.The authors claim that segmentation with SAM2 is necessary based on their ablation study. However, this may simply reflect limitations of DeepLabCut tracking. Could SAM2 be omitted if a stronger pose estimation model were used, or if the pose estimator were trained with more data?
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not provide sufficient information for reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
1.There is a duplicated “estimating” in the fourth paragraph of the Introduction. 2.In Fig. 1, the panel corresponding to Sec. 3.3 shows counting a local maximum with the minimum of two local minima. However, Sec. 3.3 states that it is counted using the maximum of two local minima. 3.Please make all notations consistent. For example, in Sec. 3.1, $I_i$ should be $\mathbf{I}_i$. 4.Please ensure that all subtitles follow a consistent style (e.g., capitalize the first letter of each subtitle).
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The overall pipeline is straightforward and easy to implement and may have practical value. However, the filtering hyperparameters, which appear to be critical in this work, may be highly data-dependent. The authors do not provide practical guidelines for selecting these parameters (e.g., based on data distributions), nor do they define them in reproducible physical units (e.g., cm instead of pixels). Due to these limitations, I suggest a weak rejection.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Reject
- [Post rebuttal] Please justify your final decision from above.
The authors’ rebuttal mostly addresses my concerns. One primary issue remains: the need for pilot experiments to determine appropriate thresholds. However, I acknowledge that this is a limitation in true real-world deployment. If the paper were categorized under the “Clinical Translation” track with an “Application” contribution, I would suggest acceptance. However, the authors position their work in the “MIC” track with a “Methodology” contribution. In this context, the novelty concern raised by Reviewer 2 remains unresolved and cannot be fundamentally addressed in the rebuttal, as the work is primarily application-oriented rather than methodological. Given the authors’ chosen positioning, I would recommend rejection due to the lack of a MIC-level methodological contribution.
Author Feedback
We are pleased that the reviewers recognized the potential and practical significance of our work, including its “high translational impact” (R1), a “meaningful problem” with “encouraging validation against accelerometer data” (R2), and its “practical value” (R3). We address primary concerns below. 1.GT Justification and Method Limitations (R1, R3): Accelerometer PSD is validated as a preclinical standard [4]. Marker-based capture [11] requires implantation, making it incompatible with our non-invasive premise. Lighting robustness was verified across conditions, maintaining r≈0.8 even under reduced brightness. 2.Novelty (R2): Our novelty lies in the problem formulation itself. No prior framework has addressed markerless, non-invasive quantification of anti-gravity tremor in freely moving mice. To solve this, we proposed two novel components absent from prior work: the Move Detection Module and prominence-based Tremor Score Estimation. Prior work either requires specialized piezoelectric sensors [1] or fundamentally fails because overhead tracking cannot capture anti-gravity vertical head oscillation [16]. In contrast, our method achieves dose-dependent replication (Figs. 3, 4), satisfying the application impact criterion under MICCAI’s joint evaluation framework. 3.Limited Dataset (R2, R3): Our dataset with n=28 (56 sequences) meets or exceeds those of prior work [1, 4, 16] (n=36, 11, 9), noting prior work [4] also uses a single fixed configuration. Thresholds were set on a held-out pilot subset, and Fig. 2 confirms wide-range stability. 4.Human Evaluation (R2, R3): Please note that human evaluation serves only as a supplementary reference, since human evaluators tend to anchor on peak intensity and overscore subsiding tremor sequences, whereas our method avoids this bias. Our primary claim is the strong correlation between our tremor score and accelerometer PSD (r=0.86). We acknowledge the single-rater limitation and plan to conduct multi-rater analysis. 5.Additional Statistical Report (R2): These details were omitted due to space. 95% bootstrap CIs (Fig. 4): SAL 10.2 [6.0, 14.5]; 5 mg/kg 27.9 [15.3, 46.4]; 10 mg/kg 58.7 [32.9, 83.4]; 20 mg/kg 75.1 [55.1, 95.0]. Non-overlapping CIs between SAL and 5mg/kg confirm lowest-dose discrimination. For r=0.86 (n=28): 95% CI≈[0.72, 0.93] (Fisher’s z). Our method is significantly superior to the baseline [16] (p<0.001). The difference vs. human (r=0.82, p≈0.63) is non-significant, confirming comparable accuracy. 6.Reproducibility and Threshold Guidelines (R3): Our recording uses 1280×720 resolution at a camera-to-subject distance that yields 1px = 0.2mm. Since this ratio depends on camera-to-subject distance for a fixed lens, it can be re-established by placing a known-size reference object in the field of view. Full setup details will be stated in the camera-ready version, with code released upon acceptance. Thresholds were not set arbitrarily but grounded in the observed data structure, as pilot experiments revealed bimodal prominence distributions with a natural valley, and both thresholds were placed at this valley. 7.Comparison Fairness with Methods Using Different Viewpoints (R3): Prior work [16] adopts an overhead view designed for tracking general locomotion, which fundamentally cannot capture the primary tremor phenomenon. Harmaline tremor manifests as anti-gravity vertical head oscillation (Fig. 5), which is inherently invisible from a top-down viewpoint. The result of r=−0.27 also supports this. 8.Ablation Study (R3): Without the Move Detection Module, vigorous locomotion contaminates tremor scores at low doses. Without SAM2 masking the background, DLC misidentifies cage elements as keypoints regardless of pose estimator strength. Multi-view ablation (r=0.82 without back camera) confirms complementary contribution. 9.Misc. (R1, R2, R3): Ethics approval and all minor corrections (Fig. 1, segmentation wording, typos, notations) will be reflected in the camera-ready.
Meta-Review
Meta-review #1
- Your recommendation
Invite for Rebuttal
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
The paper is interesting and provides novel applications description. The authors should better justify the choices and the assumptions of the methodology and improve the robustness assessment of the findings.
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
The authors have provided the requested modifications and improvements. The methodological approach is rigorous and findings impactful for the scientific community.
Meta-review #2
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Reject
- Please justify your recommendation.
The paper builds a framework for markerless motion detection and tremor severity assessment. The reviewers agreed that the application is novel and has its merit in potential clinical translation. However, reviewers raised concerns regarding the methodological novelty expected for a MIC-track contribution. The methodology-wise novelty seems limited, as the pipeline appears to be an engineering design that concatenates multiple existing components (e.g., SAM2, DeepLabCut, threshold-based filtering, and prominence-based scoring) into an application-specific framework, and the conceptual advance over prior work was viewed as incremental. While the rebuttal clarified several implementation and evaluation details, it did not substantially change this concern.
One of the reviewers suggested acceptance under the “Clinical Translation” track with an “Application” contribution since the work is motivated by potential translational relevance. However, given the large domain gap between mouse motion and human motion, there might be more pre-clinical human motion simulation studies required to further demonstrate its potential for real clinical translation. The effectiveness of the system was mainly shown for preclinical animal studies rather than real clinical deployment or near-translational capability. Thus, the AC cannot recommend acceptance under either track.
Meta-review #3
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
This is an interesting paper where the authors have proposed a framework for estimating tremor in a Parkinson’s Disease mouse model using different modalities - RGB cameras, animal pose estimation, and segmentation masks to distinguish between voluntary movement and tremor.
The reviewers have recognized the strong clinical applicability of this work. Although the developed framework is not completely innovative and is more of an engineered work to combine several different components, it is still very well designed and supported by enough experiments. The main contribution is about the unique application that has a high chance of generating interest within the MICCAI community. The paper is well written and easy to follow, thus I recommend acceptance at this stage.