Pith. sign in

REVIEW 4 major objections 4 minor 13 references

Towards Scalable Newborn Screening: Automated General Movement Assessment in Uncontrolled Settings

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A fully automatic video pipeline can predict newborn general-movement quality in uncontrolled home and hospital recordings, reaching an AUROC of up to 0.8283.

desk verdict The headline AUROC is a selected maximum from one manual labeller, not the aggregate pipeline; the paper is a useful feasibility study with fixable reporting flaws. read the letter →

arxiv 2411.09821 v4 pith:XFA6EMUT submitted 2024-11-14 cs.LG cs.CV

Daphné Chopard , Sonia Laguna , Kieran Chin-Cheong , Annika Dietz , Anna Badura , Sven Wellmann , Julia E. Vogt This is my paper · ORCID
classification cs.LGcs.CV
keywords generalmovementassessmentnewbornscreeninginfantposeestimationkeypointtrackingtime-seriesclassificationvideo-basedfidgetymovementswrithing
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

This paper sets out to show that automated General Movement Assessment (GMA) can work on real-world infant videos recorded in uncontrolled hospital and home settings, not just in fixed lab setups. Its central claim is that infant motor quality can be predicted from such videos with an AUROC of up to 0.8283, with markedly better results for the younger writhing-movement group than for the older fidgety-movement group. That would matter because GMA is a reliable early predictor of neurodevelopmental conditions, yet few clinicians are trained to score it; a video-only pipeline could make newborn screening scalable. The authors test a two-branch keypoint pipeline followed by simple time-series classifiers on 76 heterogeneous recordings, each carrying one coarse movement-quality label per video.

What carries the argument

The load-bearing machinery is a two-branch keypoint-extraction pipeline that converts raw video into pose time series. The Label & Track branch seeds 17 anatomical keypoints in one frame by hand and tracks them with the TAPIR point tracker, with automatic outlier detection prompting manual correction; the AggPose branch labels 21 keypoints in every frame independently using a vision transformer for infant pose estimation. From the tracked keypoints the pipeline computes x-y coordinate channels and joint angles at selected triplets (e.g., elbow angle, knee angle, head-to-shoulder angle), then feeds these time series to a 1D-CNN, an LSTM, or a Random Forest. The key mechanism is that angle features and temporal tracking encode movement quality without requiring frame-level movement annotations.

What would settle it

Re-annotate each video at segment level, marking which 15–30 second windows actually contain informative writhing or fidgety movement, retrain the same classifiers on clip-level labels, and compare AUROC; if clip-level performance falls well below the reported 0.8283, the headline result is carried by video-level label sharing rather than by real movement signal.

Watch

Extended reading notes

Core claim

The paper's central discovery, stated on its own terms, is that coarse video-level annotations and pose-derived time series are sufficient to predict general-movement quality: the best configuration reaches an AUROC of 0.8283, and the early-GM group (preterm infants in the writhing phase) is classified substantially better than the late-GM group (infants in the fidgety phase). Across classifiers and keypoint-extraction methods, angle-based features, alone or combined with coordinates, generally outperform raw coordinate channels, and the manual-labelling-plus-tracking branch performs comparably to, and in several settings better than, fully automatic per-frame pose estimation. The authors interpret this as evidence that automated GMA is feasible in uncontrolled, real-world settings when the preprocessing pipeline handles resampling, cropping, outlier correction, and stratified splitting.

Load-bearing premise

The load-bearing assumption is that every fixed-length clip cut from a video inherits the video's single movement-quality label, even though only part of a recording may show diagnostically informative movement, and the paper itself concedes in its limitations that not all segments are equally informative.

Editorial extensions

If this is right

  • A screening tool could be assembled from equipment parents already have: any home video, one manual label in a single frame (or none with automatic labelling), and a light-weight classifier.
  • Angle-based representations of body configuration should be treated as a primary feature for movement-quality classification, since they consistently beat raw coordinates in the reported experiments.
  • Because the early-GM group performs much better than the late-GM group, age-specific models or age-specific features will likely be needed before fidgety-movement screening is reliable.
  • Simple, compute-efficient models (Random Forest, 1D-CNN, LSTM) are competitive on this task, so scalable deployment need not wait for large transformer models.

Reading between the lines

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

  • Segment-level labels would likely move the reported numbers: if informative 15–30 second windows were identified, AUROC should either rise by removing noise or fall, revealing that the video-level label was carrying the signal; re-labelling a subset of clips would settle which.
  • The late-GM gap suggests fidgety-movement quality may be more temporal in nature, so sustained-motion features (e.g., velocity or wiggle statistics over seconds) might close the gap more than per-frame angles.
  • The high-risk cohort (about 65% and 27% abnormal in the two groups) means the AUPRC values are not transferable to general newborn screening; evaluating at roughly the 3% population prevalence is the decision-relevant check.
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 / 4 minor

Summary. The paper presents a pipeline for automated General Movement Assessment (GMA) from infant videos recorded in uncontrolled settings. The pipeline extracts keypoints either by manual labelling in one frame plus TAPIR tracking (Label & Track) or fully automatically with AggPose, preprocesses the videos (resampling, cropping, outlier correction, fixed-length clipping), computes coordinate and angle features, and trains 1D-CNN, LSTM, and random forest classifiers. The evaluation uses a dataset of 76 infants split into early and late GM groups, with video-level binary labels, and reports AUROC, AUPRC, and accuracy under 5-fold stratified cross-validation. The central claim, stated in the conclusion, is that infant motor quality can be predicted from video recordings with an AUROC of up to 0.8283, supporting the feasibility of automated GMA in real-world settings.

Significance. If the central claim holds, this is a useful preliminary contribution toward scalable newborn screening: it demonstrates that a combination of existing pose estimation/tracking tools and simple time-series classifiers can operate on heterogeneous, clinically realistic video data. The paper has concrete strengths: the code is publicly available, the dataset is genuinely uncontrolled (varied devices, settings, ages), the comparison between manual tracking and fully automatic pose estimation is informative, and the limitations are acknowledged candidly, including the small dataset, label noise, and demographic homogeneity. However, the headline number is not produced by the main evaluation protocol, and the experimental reporting is not yet precise enough to support the strength of the conclusion. As a preliminary benchmark with a clearly qualified performance estimate, the work is valuable; in its current form, the central feasibility claim is overstated relative to what the tables show.

major comments (4)
  1. [§4.1] The hyperparameter selection protocol is described as: "We performed a grid search over the hyperparameter space for each of the three classifier types in Section 3.4, choosing the values that performed best in the evaluation. The final models were trained using these hyperparameters with 5-fold stratified CV, and the results from the held-out test set, with 20% of the data, were collected." This is ambiguous and potentially leaky: if "the evaluation" refers to the same held-out test set, then the reported numbers are optimistically biased because hyperparameters were selected on the test data. If a separate validation split was used, that split is not described. Please specify the exact selection procedure, including whether any portion of the test set was used during model selection, and if so, report results from a truly held-out split or use nested cross-validation. This concern affects every number in Tables 3-6.
  2. [§5 and Appendix B, Table 4] The conclusion states "We demonstrate that infant motor quality can be predicted from video recordings with an AUROC of up to 0.8283," but this value traces to a single configuration in Table 4: Early GM, Labeller #1, Angles, Random Forest (0.83 ± 0.13). It is not the aggregate result of the pipeline (Table 3 shows Early GM Label & Track Angles RF at 0.77, and fully automatic AggPose peaks at 0.72), and it is selected as the maximum over three manual labellers, three models, three feature sets, and two age groups with no multiple-comparison control. Moreover, this best configuration requires per-video manual keypoint labelling and two rounds of manual outlier correction, so it does not support the "fully automatic" or "scalable" reading of the claim. The central claim should be re-anchored to the aggregate or fully automatic results, or the "up to" number must be clearly labelled as the best manual-labeller scenario with appropriate caveats.
  3. [§3.2 and §4.2] The effective sample size is 76 infants, but the evaluation treats each fixed-length clip as a separate instance. If a single video produces multiple clips, those clips share the same video-level label and are highly correlated; the reported standard deviations are over 30 seeds only, not over infants or videos. The 20% held-out test set therefore very likely contains clips, not 20% of infants, even though the text claims "no infant appears in both the training and testing datasets." Please clarify whether the split is performed at the infant level or the clip level, report the number of clips per infant, and consider computing performance with infant-level bootstrapping or with a single clip per infant to quantify the true variability. Without this, the reported confidence intervals and the generalizability claim are difficult to assess.
  4. [§3.2 and §5] The label-noise issue is load-bearing: every fixed-length clip receives the video's single binary label, even though the paper itself acknowledges in §5 that "not all segments of a video are equally informative." This means the training targets are partially wrong, and the reported AUROC may reflect classification of uninformative segments as often as informative ones. The paper mentions this as a limitation but does not quantify the noise or test robustness, e.g., by training on a subset of clips with higher motion, by using multiple instance learning, or by annotating segment-level labels for a subsample. Since the central claim is about feasibility of automated screening, the effect of this label noise on the reported numbers should be at least qualitatively assessed.
minor comments (4)
  1. [Throughout] Typos and misspellings: "bottleck" should be "bottleneck" (§4.1), "sptial" should be "spatial" (§5), "Addionally" should be "Additionally" (§5), "T rack" should be "Track" (§3.1), and "F uture" should be "Future" (§5).
  2. [Table 3 caption] The caption says "The mean performance across 30 seeds and 3 keypoint labellers is reported for the Label & Track method," but Table 3 reports only point estimates without standard deviations for Label & Track, whereas Appendix B Tables 4-6 include standard deviations. Please report standard deviations in Table 3 as well, or indicate that they are deferred to the appendix.
  3. [§3.2] The description "Extreme keypoints are then labelled" is confusing because the previous paragraph already discusses "extreme keypoints labelled" in the labelling step. Please clarify the distinction between the keypoint labelling step and the later use of extreme keypoints for cropping.
  4. [Figure 4] The list of keypoints in Figure 4 is hard to read because the "All Keypoints" and "Extreme" columns are not visually separated; the x markers make it unclear which keypoints are extreme. Consider reformatting this table for readability.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the classifier evaluation is self-contained; the flagged issues are statistical reporting and validation-protocol concerns, not definitional reductions.

full rationale

The paper's central claim, that infant motor quality can be predicted from video recordings with an AUROC of up to 0.8283, is an empirical evaluation result rather than a derivation from first principles. The direction of inference is standard and non-circular: clinically assigned binary labels (from external criteria in Einspieler and Prechtl, 2005), keypoint trajectories (TAPIR, AggPose, both third-party external tools), hand-specified angle features (based on Prakash et al., 2023), and standard classifiers (CNN, LSTM, RF) are combined, with 5-fold stratified CV and a held-out 20% test split, and metrics are reported on the held-out data. The paper contains no self-citations that are load-bearing: the references to TAPIR, AggPose, Einspieler, and others are all external sources with independent existence, and no 'uniqueness theorem' or prior-work ansatz from the authors is invoked to force the modelling choices. The reader's skeptical concerns are real but belong to soundness and reporting, not circularity: the headline 0.8283 appears only as Labeller #1's Early-GM angle/RF result in Appendix B (Table 4), not in the main aggregate Table 3, so the 'up to' phrasing reflects best-of selection over 3 labellers × 3 models × 3 feature sets × 2 age groups without multiple-comparison control; and the grid-search description ('choosing the values that performed best in the evaluation') is under-specified as to which split was used for selection. These concerns could inflate the reported number, but they do not make the prediction equal to its inputs by construction, and the manuscript itself honestly discloses the coarse-label limitation ('not all segments of a video are equally informative') in Section 5. No fitted parameter is renamed as a prediction, no known result is renamed, and no definitional equivalence between inputs and outputs is exhibited. Accordingly, the finding is no significant circularity, with a minor score reflecting only the under-described selection protocol as a transparency caveat.

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

The central claim relies on a chain of untested-in-this-cohort assumptions: the keypoint extractors work on this heterogeneous data, the video-level labels transfer to all clips, and the geometric features suffice. The only fitted quantities are pipeline choices and hyperparameters, which are not framed as scientific predictions. No new entities are introduced.

free parameters (5)
  • Outlier detection threshold = 15 times overall standard deviation per keypoint
    Hand-chosen threshold in Section 3.2 to flag unrealistic keypoint jumps; not derived from a performance objective.
  • Outlier correction rounds = 2 rounds
    Set to two after 'marginal improvements' after three, a heuristic choice in Section 3.2.
  • Fixed-length clip duration = 616 frames (early), 674 frames (late), about 20.5 s and 22.5 s
    Determined by the shortest recording in each age group (Section 3.2), a data-dependent choice that affects all downstream analysis.
  • Crop margin = 15% margin
    Hand-chosen normalization factor in Section 3.2 to keep the infant consistently framed.
  • Model hyperparameters = e.g., CNN LR=0.00001, batch=6, 150 bottleneck features, 150 epochs; LSTM LR=0.001, batch=6, 3 layers, hidden=64, 200…
    Selected by grid search in Section 4.1; not held out by a separate validation split, so they are fit to the data.
assumptions (5)
  • domain assumption TAPIR accurately tracks the manually labelled keypoints across heterogeneous infant video frames, including occlusions.
    Relied on in Section 3.1.1; if tracking fails or errors correlate with movement quality, all downstream features are corrupted.
  • domain assumption AggPose provides accurate per-frame keypoint positions for infants in this dataset.
    Used as the automatic labelling baseline in Section 3.1.2; its training data may not match this clinical cohort.
  • domain assumption The binary video-level GM label assigned by the two physicians is valid for every fixed-length clip extracted from that video.
    The paper itself notes in Section 5 that only part of a video may reflect abnormal movements, so clip labels inherit noise.
  • domain assumption Coordinates and angles between the selected keypoints retain the discriminative information needed for GM classification.
    Feature choice in Section 3.3; if movement quality is encoded in dynamics not captured by these geometric features, the approach fails.
  • domain assumption Stratified splits at the infant level are sufficient to prevent train and test information leakage.
    Stated in Section 3.2; however, multiple clips from the same video still share camera setup and infant identity, and the effective sample size is 76.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Scalable Newborn Screening: Automated General Movement Assessment in Uncontrolled Settings." pith.science (2026). https://pith.science/paper/XFA6EMUT

@misc{pith2026241109821,
  author       = {Pith},
  title        = {Pith review of: Towards Scalable Newborn Screening: Automated General Movement Assessment in Uncontrolled Settings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XFA6EMUT}},
  note         = {Machine review of arXiv:2411.09821}
}
read the original abstract

General movements (GMs) are spontaneous, coordinated body movements in infants that offer valuable insights into the developing nervous system. Assessed through the Prechtl GM Assessment (GMA), GMs are reliable predictors for neurodevelopmental disorders. However, GMA requires specifically trained clinicians, who are limited in number. To scale up newborn screening, there is a need for an algorithm that can automatically classify GMs from infant video recordings. This data poses challenges, including variability in recording length, device type, and setting, with each video coarsely annotated for overall movement quality. In this work, we introduce a tool for extracting features from these recordings and explore various machine learning techniques for automated GM classification.

Figures

Figures reproduced from arXiv: 2411.09821 by the authors.

Figure 1
Figure 1. Histograms showing the distributions of (a) video length, (b) frame rate per second (FPS), and (c) video resolution for each GM group. The videos are highly heterogeneous, with substantial variation in recording length, frame rate, and resolution. FPS values are rounded to the closest one. 3. Methods This work involves extracting features from specific anatomical landmarks of interest— so-called keypoints (KP)—in vi… view at source ↗
Figure 2
Figure 2. Overview of the GMA classification pipeline. Video recordings undergo prepro￾cessing steps (Section 3.2). Keypoints are then extracted using one of two ap￾proaches described in Section 3.1: (a) Label & Track, or (b) AggPose. Fea￾tures are extracted (Section 3.3) from the keypoints and used to train GMA classification models (Section 3.4) including CNN, LSTM, and Random Forest for prediction. tracking, forms the back… view at source ↗
Figure 3
Figure 3. Consecutive frames with unrealistic keypoint tracking. The outlier detection tool displays potential unrealistic point movements and allows manual correction. 3.1.2. AggPose: Automatic label As a benchmark, we also evaluate a fully automated keypoint extraction method: AggPose (Cao et al., 2022), a vision transformer model specifically designed for infant pose estimation and trained on a large-scale infant pose data… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: List of all and extreme keypoints considered in this study for manual la￾belling, coloured by label on the right. All Key￾points Extreme nose head bot￾tom head top x left ear right ear left shoulder right shoul￾der left elbow x right elbow x left wrist x right wrist x …
Figure 6
Figure 6. Figure 6: Illustration of the angle fea￾tures. The dash lines show the angles that we compute in the set of angle features. Keypoint Coordinates The x- and y-coordinates of all labelled keypoints are treated in￾dependently as separate input channels, and used as a simple represe…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 11 canonical work pages

  1. [1]

    Movidea: A software package for automatic video analysis of movements in infants at risk for neurodevelopmental disorders

    Walter Baccinelli, Maria Bulgheroni, Valentina Simonetti, Francesca Fulceri, Angela Caruso, Letizia Gila, and Maria Luisa Scattoni. Movidea: A software package for automatic video analysis of movements in infants at risk for neurodevelopmental disorders. Brain sciences, 10 0 (4): 0 203, 2020

    work page 2020

  2. [2]

    A random forest guided tour

    G \'e rard Biau and Erwan Scornet. A random forest guided tour. Test, 25: 0 197--227, 2016

    2016

  3. [3]

    Random forests

    L Breiman. Random forests. Machine Learning, 2001

    work page 2001

  4. [4]

    Tap-vid: A benchmark for tracking any point in a video

    Carl Doersch, Ankush Gupta, Larisa Markeeva, Adria Recasens, Lucas Smaira, Yusuf Aytar, Joao Carreira, Andrew Zisserman, and Yi Yang. Tap-vid: A benchmark for tracking any point in a video. Advances in Neural Information Processing Systems, 35: 0 13610--13626, 2022

    work page 2022

  5. [5]

    Prechtl's assessment of general movements: a diagnostic tool for the functional assessment of the young nervous system

    Christa Einspieler and Heinz FR Prechtl. Prechtl's assessment of general movements: a diagnostic tool for the functional assessment of the young nervous system. Mental retardation and developmental disabilities research reviews, 11 0 (1): 0 61--67, 2005

    work page 2005

  6. [6]

    Long short-term memory

    Sepp Hochreiter and J \"u rgen Schmidhuber. Long short-term memory. Neural computation, 9 0 (8): 0 1735--1780, 1997

    1997

  7. [7]

    Ai approaches towards prechtl’s assessment of general movements: A systematic literature review

    Muhammad Tausif Irshad, Muhammad Adeel Nisar, Philip Gouverneur, Marion Rapp, and Marcin Grzegorzek. Ai approaches towards prechtl’s assessment of general movements: A systematic literature review. Sensors, 20 0 (18): 0 5321, 2020

    work page 2020

  8. [8]

    1d convolutional neural networks and applications: A survey

    Serkan Kiranyaz, Onur Avci, Osama Abdeljaber, Turker Ince, Moncef Gabbouj, and Daniel J Inman. 1d convolutional neural networks and applications: A survey. Mechanical systems and signal processing, 151: 0 107398, 2021

    work page 2021

Show all 13 references
  1. [9]

    Video-based real-time assessment and diagnosis of autism spectrum disorder using deep neural networks

    Varun Ganjigunte Prakash, Manu Kohli, Aragulla Prasad Prathosh, Monica Juneja, Manushree Gupta, Smitha Sairam, Sadasivan Sitaraman, Anjali Sanjeev Bangalore, John Vijay Sagar Kommu, Lokesh Saini, et al. Video-based real-time assessment and diagnosis of autism spectrum disorder...

    2023

  2. [10]

    Automated movement analysis to predict cerebral palsy in very preterm infants: an ambispective cohort study

    Kamini Raghuram, Silvia Orlandi, Paige Church, Maureen Luther, Alex Kiss, and Vibhuti Shah. Automated movement analysis to predict cerebral palsy in very preterm infants: an ambispective cohort study. Children, 9 0 (6): 0 843, 2022

    2022

  3. [11]

    o lte, Karin Nielsen-Saines, Florian B Pokorny, Robert Peharz, Luise Poustka, Florentin W \

    Simon Reich, Dajie Zhang, Tomas Kulvicius, Sven B \"o lte, Karin Nielsen-Saines, Florian B Pokorny, Robert Peharz, Luise Poustka, Florentin W \"o rg \"o tter, Christa Einspieler, et al. Novel ai driven approach to classify infant motor functions. Scientific Reports, 11 0 (1): ...

    2021

  4. [12]

    General movement assessment by machine learning: why is it so difficult? Journal of Medical Artificial Intelligence, 2 0 (July): 0 15, 2019

    William Thomas Schmidt, Matthew Regan, Michael C Fahey, and Andrew Paplinski. General movement assessment by machine learning: why is it so difficult? Journal of Medical Artificial Intelligence, 2 0 (July): 0 15, 2019

    2019

  5. [13]

    The future of general movement assessment: The role of computer vision and machine learning--a scoping review

    Nelson Silva, Dajie Zhang, Tomas Kulvicius, Alexander Gail, Carla Barreiros, Stefanie Lindstaedt, Marc Kraft, Sven B \"o lte, Luise Poustka, Karin Nielsen-Saines, et al. The future of general movement assessment: The role of computer vision and machine learning--a scoping revi...

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.