Pith. sign in

REVIEW 2 major objections 7 minor 32 references

A motion descriptor built from deformable image registration detects five cardiac keyframes in cine CMR and beats left-ventricular volume curves for end-diastole and end-systole.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 11:14 UTC pith:4WQFSUTZ

load-bearing objection Solid SAX evidence and a useful five-keyframe extension, but the §2.3 rule equations are circular as written and the 4CH abstract claim overreaches — fixable before this should be trusted. the 2 major comments →

arxiv 2510.05819 v1 pith:4WQFSUTZ submitted 2025-10-07 cs.CV cs.AI

Deformable Image Registration for Self-supervised Cardiac Phase Detection in Multi-View Multi-Disease Cardiac Magnetic Resonance Images

Sven Koehler , Sarah Kaye Mueller , Jonathan Kiekenap , Gerald Greil , Tarique Hussain , Samir Sarikouch , Florian Andr\'e , Norbert Frey
show 1 more author
Sandy Engelhardt
This is my paper
classification cs.CV cs.AI
keywords cardiac keyframe detectiondeformable image registrationself-supervised learningmotion descriptorcine CMRcardiac magnetic resonance imagingend-diastole and end-systolemulti-view cardiac imaging
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper proposes that cardiac phase can be read directly from the direction of myocardial motion rather than from changes in ventricular volume. It builds dense displacement fields between consecutive cine frames using an unsupervised deformable registration network, then compresses each field into a single curve that is negative while the heart contracts and positive while it relaxes. From zero-crossings and extrema of this curve, the method flags five keyframes: end-diastole, mid-systole, end-systole, peak filling, and mid-diastole. Across several multi-centre, multi-disease datasets plus a rare congenital-heart-defect cohort, the method detects ED and ES with 30–51% lower error than a supervised volume-based baseline in short-axis views, and it still finds the extra keyframes within about 1.3 frames in short-axis and 1.7 frames in long-axis views. If these results hold, cardiac MRI analysis could be temporally aligned across patients and cycles without manual labels or ECG.

Core claim

The central claim is that myocardial displacement fields, not chamber-volume curves, are the right signal for timing the cardiac cycle. The paper shows this by defining a one-dimensional descriptor that averages, over a motion-masked region, the cosine similarity between each voxel's displacement and a vector pointing toward a focus point; negative values mean contraction, positive values mean relaxation. The smoothed descriptor reliably crosses zero at end-diastole and end-systole, and its global minimum plus derivative extrema identify mid-systole, peak filling, and mid-diastole. On large multi-centre, multi-disease evaluations, this self-supervised rule-based method detects ED and ES with

What carries the argument

The carrying object is the deformable motion descriptor, a 1D curve computed from the dense vector field produced by a convolutional registration network. Per voxel, the cosine similarity is taken between the displacement vector and a position vector pointing from that voxel to a focus point; a rule-based mask keeps only voxels with persistent large motion and strong directional change, and the descriptor is the masked mean. The focus point is computed self-supervised as the temporal mean centre of mass of the masked field, so the whole pipeline needs no labels. A simple rule set converts the curve's shape into keyframes: the global minimum is mid-systole, zero crossings with appropriate slo

Load-bearing premise

The method assumes the automatically chosen focus point sits inside the left ventricle, so that motion toward it means contraction; in several long-axis cases it lands in the atria, and then the sign of the descriptor no longer cleanly separates contraction from relaxation.

What would settle it

A controlled experiment that fixes the registration model but perturbs the focus point across chamber boundaries would settle the load-bearing assumption: if keyframe error stays flat when the focus point moves outside the left ventricle, the interpretable contraction/relaxation sign is not what is carrying the accuracy; if error rises, the assumption is confirmed.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Keyframe detection becomes possible without manual labels or ECG, using only cine images and a registration model trained on unlabelled frames.
  • End-diastole and end-systole are found within about one frame in short-axis cine MRI across unseen centres, comparable to inter-observer variability, making automated temporal alignment practical.
  • The same one-dimensional curve yields three additional physiologically meaningful keyframes, enabling phase-resolved analysis beyond the traditional ED–ES pair.
  • Because the registration is self-supervised and the phase rules are fixed, the method transfers to unseen scanners, pathologies, and a rare congenital heart defect cohort without retraining.
  • Replacing volume curves with motion curves addresses a known blind spot: iso-volumetric contraction and relaxation change myocardial motion without changing chamber volume.
  • The approach enables inter- and intra-patient comparison of cardiac dynamics at aligned phases regardless of cycle length or starting phase.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the continuous descriptor curve could serve as a cardiac-phase clock, so any intermediate phase could be interpolated from its shape rather than limited to five named keyframes.
  • Editorial extension: if atrial/ventricular motion overlap is the main cause of weaker long-axis results, separate focus points for ventricles and atria could isolate chamber-specific contraction and improve accuracy.
  • Editorial extension: the same descriptor idea may transfer to other dynamic imaging modalities with a contracting focus, such as echocardiography, where the smaller field of view would need a different focus-point strategy.
  • Editorial extension: because the masking thresholds were chosen empirically, a per-patient adaptive threshold is a direct testable extension for low-resolution or artefact-heavy acquisitions.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 7 minor

Summary. This paper presents a fully self-supervised method for detecting five cardiac keyframes (end-diastole, mid-systole, end-systole, peak flow, mid-diastole) in short-axis (SAX) and four-chamber (4CH) cine CMR. A deformable registration U-Net estimates dense displacement fields; a 1D motion descriptor α_t is computed by masking the field and aggregating voxel-wise cosine similarities between displacement vectors and position vectors pointing to a focus point. Five keyframes are detected from α_t via a rule set based on zero crossings and extrema. The method is evaluated on M&Ms-2, M&Ms, ACDC, and a tetralogy-of-Fallot cohort (GCN), and compared against a supervised LV-volume-based baseline. The central hypothesis is that motion-based phase detection outperforms volume-based detection.

Significance. If the results hold, the work would provide a label-free, scanner-generalizable alternative to volume-based cardiac phase detection, with additional keyframes enabling phase-aligned analysis. Its strengths include evaluation on multiple public multi-center, multi-disease datasets, including rare congenital heart disease, and the stated intent to release code and annotations. However, two issues currently block acceptance: the keyframe rule set is not an executable specification as written, and the abstract's 4CH improvement claim is not supported by the fully self-supervised results.

major comments (2)
  1. [§2.3] The keyframe rule set is circular and not executable as written. ES is defined as max{α(t)=0, α′(t)>0} over t∈[MS, PF], while PF is simultaneously defined as min{α′(t)=0, α″(t)<0} over t∈[ES, MS]. ED depends on PF and MD depends on ED. Therefore each equation references a keyframe that is itself defined by the same rule set, so the system cannot be evaluated without already knowing the solution. The text says the rules are applied as a sequence to the cyclic sub-sequence, but the ordering and cyclic-index convention are not stated. Since the reported cFD results in Tables 3–4 are produced by this rule set, the method is not reproducible from the manuscript alone. Please replace the equation block with an explicit, executable algorithm, e.g., pseudo-code that specifies the detection order (e.g., locate MS, then ES as the first zero crossing after MS, then PF as the first maximum after ES,
  2. [Abstract / Table 4] The abstract claims 'improved detection accuracy by 30%–51% for SAX and 11%–47% for 4CH in ED and ES ... compared with the volume-based approach.' For the fully self-supervised focus point C_mse, Table 4 shows no improvement over the LV-volume baseline on M&Ms-2 test (ED: 0.94±1.11 vs 0.93±1.20; ES: 0.99±1.09 vs 0.91±1.20) and only marginal improvements on M&Ms-2 train. The 11% lower bound of the 4CH range does not appear to be supported by any C_mse result; it may stem from segmentation-derived focus points such as C_sept or C_lv. This overstatement concerns the central claim that the motion-descriptor method is more accurate than volume-based detection. Please restrict the summary claim to the specific focus-point variant used, or report the ranges separately for fully self-supervised and segmentation-derived focus points.
minor comments (7)
  1. [Header] The preprint header contains a typo: 'Medical Imaga Analysis' should be 'Medical Image Analysis'.
  2. [§3 / Table 3] There is an inconsistency in p-value reporting. The text says for M&Ms ED p<0.1e-3, but Table 3 shows '**' (p<0.01). Please standardize the reporting.
  3. [Table 4] The note 'In case of the M&Ms-2 test dataset, the mean cFD would be 0.82±0.89 if only ED and ES are considered like it is the case for base' does not state which focus point this refers to. Please specify (presumably C_mse) and clarify the comparison.
  4. [Discussion] The statement that the rule set 'has been further optimized to achieve optimal performance for healthy hearts' is not accompanied by a description of the optimization procedure, the data used, or whether labels were consulted. This is important for reproducibility and for assessing potential bias toward healthy subjects. Please describe the tuning process.
  5. [Abstract] The terms 'LAX' and '4CH' are used interchangeably (Abstract uses 'LAX', rest of the paper uses '4CH'). Please unify.
  6. [§2.5 Eq. (12)] The cFD formula can be simplified to min(|p_i−p̂_i|, T−|p_i−p̂_i|). Consider using this more readable form.
  7. [§2.4 / Tables] The number of keyframe annotations per dataset and the inter-observer variability for the additional keyframes (MS, PF, MD) are not fully reported. Please add this information, as it is essential for judging the reliability of the 'all keyframes' results.

Circularity Check

2 steps flagged

Keyframe rule set and self-supervised focus point are self-referential as written, making the method's derivation chain formally circular despite independent empirical evaluation.

specific steps
  1. self definitional [Section 2.3, Cardiac Keyframes (rule equations after MS definition)]
    "ES=max{α(t) = 0 and α′(t)>0} for t∈[MS, PF]; PF=min{α′(t) = 0 and α′′(t)<0} for t∈[ES, MS]; ED=max{α(t) = 0 and α′(t)<0} for t∈[PF, MS]; MD=max{α′(t) = 0 and α′′(t)<0} for t∈[PF, ED]"

    The rule for ES searches only within [MS, PF], so PF must already be known before PF is computed; the rule for PF searches only within [ES, MS], so ES must already be known. ED similarly depends on PF, and MD depends on both PF and ED. With the intervals read literally, every keyframe except the initial MS is defined in terms of another keyframe that is itself an output of the same rule set. The derivation chain does not terminate, and the cFD results in Tables 3–4 cannot be reproduced from this formal description unless an unstated cyclic ordering or fixed-point convention is imposed.

  2. self definitional [Section 2.6 (focus point C_mse) and Section 2.2, Eqs. (5)–(11)]
    "In our self-supervised approach, the focus point C is defined as the centre of mass of the computed mask M, averaged along the temporal axis, denoted as C_mse. ... MΔα(xi)=H(Δαi−TΔα) ... Δαi = max t (αi(t))−min t (αi(t))"

    C_mse is defined as the center of mass of the mask M, but the mask's directional component M_Δα is computed from Δα_i, which depends on α_i(t)=cos(v_i(t), C−x_i) — i.e., on the very focus point being defined. Thus C_mse satisfies the fixed-point equation C = COM[M(C)], and no initialization or iteration is described in the paper. Since every keyframe is derived from α_t built using C, the entire self-supervised pipeline is well-defined only up to this circular construction.

full rationale

The paper's empirical comparison is not statistically forced: the motion descriptor is computed from registration fields, the volume-based baseline is independent, and evaluation uses public/external ground-truth annotations. No load-bearing uniqueness theorem or ansatz is imported via self-citation; prior self-citations (Koehler et al. 2022a, 2025) support continuity but the method is described in this paper. However, two formal definitions in the derivation chain are circular as written: the Section 2.3 keyframe rules reference each other (ES needs PF while PF needs ES; ED and MD similarly), and the Section 2.6 self-supervised focus point C_mse is the center of mass of a mask that itself depends on C_mse through M_Δα. These circularities make the reported cFD values not derivable from the manuscript's equations alone without unstated conventions. The central claim still has independent empirical content, but the method description is partially self-referential, so the score is moderate rather than extreme.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

The central contribution is a derived scalar descriptor, not a new physical entity, so no invented entities are listed. However, the keyframe mapping rests on empirically tuned thresholds, a single-focus-point contraction/relaxation assumption, and physician labels whose uncertainty is not quantified for the new keyframes.

free parameters (5)
  • T_norm (magnitude percentile threshold) = 50th percentile for both SAX and 4CH
    Empirically determined mask threshold; directly controls which voxels contribute to the motion descriptor and hence to keyframe positions (§2.2, §2.6).
  • T_Δα (directional-change threshold) = 0.8 for SAX, 1.2 for 4CH
    Empirically determined threshold for temporal directional change; affects which voxels are kept in the mask and therefore the shape of α_t (§2.2, §2.6).
  • Gaussian smoothing σ = 2
    Applied to α_t before rule-based keyframe detection; authors note it can shift zero crossings by <1 frame (§2.2).
  • Registration regularization weight λ = 0.001
    Weight on the diffusion regularizer; taken from prior registration work but still a hand-set hyperparameter affecting the deformation fields (§2.1).
  • Rule-set tuning for healthy hearts = unspecified
    Discussion states the rules were "further optimized to achieve optimal performance for healthy hearts"; no objective function, search space, or validation split is reported.
axioms (5)
  • domain assumption Consecutive-frame pull-registration fields estimated by a U-Net trained with SSIM and diffusion regularization faithfully represent cardiac motion, including through-plane SAX motion and 4CH atrial/ventricular motion.
    The entire descriptor is derived from these fields; no registration-quality validation is reported beyond downstream keyframe accuracy (§2.1).
  • domain assumption A single scalar cosine similarity averaged over masked voxels relative to one fixed focus point captures global cardiac contraction and relaxation.
    This is the core modeling assumption in §2.2, Eq. (5)–(6); it is plausible but not derived from cardiac physiology.
  • ad hoc to paper Zero crossings and extrema of the smoothed α_t curve correspond one-to-one to the five physiological keyframes (ED, MS, ES, PF, MD) across healthy and pathological hearts.
    Stated as observed rules in §2.3; the mapping is validated only indirectly through manual annotations and was partly tuned for healthy hearts.
  • ad hoc to paper The magnitude and directional-change masking rules remove non-cardiac motion without removing informative myocardial motion.
    Mask thresholds T_norm and T_Δα are empirically chosen; no ablation or sensitivity analysis is reported (§2.2, §2.6).
  • domain assumption Physician annotations of the five keyframes, including the newly added MS/PF/MD labels, are reliable reference standards.
    Only ED/ES inter-observer variability is reported; no IOV is given for the three new keyframes (§2.4, Tables 3–4).

pith-pipeline@v1.3.0-alltime-deepseek · 20768 in / 13370 out tokens · 82398 ms · 2026-08-04T11:14:21.711961+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Deformable Image Registration for Self-supervised Cardiac Phase Detection in Multi-View Multi-Disease Cardiac Magnetic Resonance Images." pith.science (2026). https://pith.science/paper/4WQFSUTZ

@misc{pith2026251005819,
  author       = {Pith},
  title        = {Pith review of: Deformable Image Registration for Self-supervised Cardiac Phase Detection in Multi-View Multi-Disease Cardiac Magnetic Resonance Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4WQFSUTZ}},
  note         = {Machine review of arXiv:2510.05819}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Cardiovascular magnetic resonance (CMR) is the gold standard for assessing cardiac function, but individual cardiac cycles complicate automatic temporal comparison or sub-phase analysis. Accurate cardiac keyframe detection can eliminate this problem. However, automatic methods solely derive end-systole (ES) and end-diastole (ED) frames from left ventricular volume curves, which do not provide a deeper insight into myocardial motion. We propose a self-supervised deep learning method detecting five keyframes in short-axis (SAX) and four-chamber long-axis (4CH) cine CMR. Initially, dense deformable registration fields are derived from the images and used to compute a 1D motion descriptor, which provides valuable insights into global cardiac contraction and relaxation patterns. From these characteristic curves, keyframes are determined using a simple set of rules. The method was independently evaluated for both views using three public, multicentre, multidisease datasets. M&Ms-2 (n=360) dataset was used for training and evaluation, and M&Ms (n=345) and ACDC (n=100) datasets for repeatability control. Furthermore, generalisability to patients with rare congenital heart defects was tested using the German Competence Network (GCN) dataset. Our self-supervised approach achieved improved detection accuracy by 30% - 51% for SAX and 11% - 47% for 4CH in ED and ES, as measured by cyclic frame difference (cFD), compared with the volume-based approach. We can detect ED and ES, as well as three additional keyframes throughout the cardiac cycle with a mean cFD below 1.31 frames for SAX and 1.73 for LAX. Our approach enables temporally aligned inter- and intra-patient analysis of cardiac dynamics, irrespective of cycle or phase lengths. GitHub repository: https://github.com/Cardio-AI/cmr-multi-view-phase-detection.git

Figures

Figures reproduced from arXiv: 2510.05819 by Florian Andr\'e, Gerald Greil, Jonathan Kiekenap, Norbert Frey, Samir Sarikouch, Sandy Engelhardt, Sarah Kaye Mueller, Sven Koehler, Tarique Hussain.

Figure 1
Figure 1. Figure 1: Graphical abstract. Overview of the proposed pipeline. The top row illus￾trates the input data and the self-supervised deformable image registration model. The bottom row shows the interpretation of the resulting dense deformable vector field as motion direction, enabling derivation of a one-dimensional motion descriptor for cardiac key-frame detection. 2 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: LV volume curve, ECG signal and our proposed motion descriptor α over the cardiac cycle. The figure shows the temporal relation between cardiac phases and keyframes, left ventricular volume (top blue curve), ECG (middle red curve), and the motion descriptor α (bottom black curve) derived from CMR data. The cycle is divided into systole (blue) and diastole (red), with iso-volumetric contraction time (ICT) a… view at source ↗
Figure 3
Figure 3. Figure 3: Self-supervised rule-based masking of CMR a) Schematic illustration of computation of the direction of motion α. The motion vector ⃗v from each voxel xi is compared to a reference position vector ⃗w, which points from the corresponding voxel to a fixed anatomical focus point Cn. The angle between ⃗v and ⃗w is quantified by their cosine similarity α = cos(⃗v, ⃗w) ∈ [−1, 1]. This scalar α represents the dire… view at source ↗
Figure 4
Figure 4. Figure 4: Motion descriptor αt for all five SAX datasets. Each subplot shows the median of the masked αt (blue/left axis) with IQR (light blue/left axis) of each dataset along with its median displacement magnitude | −→v t| (red/right axis), which is normalized in a range of [0, 1]. Every input was linearly interpolated to 40 frames. The averaged phase indices (x-axis) are displayed together with the corresponding p… view at source ↗
Figure 5
Figure 5. Figure 5: Motion descriptor αt for all three 4CH datasets [bottom]. Each subplot shows the median of the masked αt (blue/left axis) with IQR (light blue/left axis) of each dataset along with its median displacement magnitude | −→v t| (red/right axis), which is normalized in a range of [0, 1]. Every input was linearly interpolated to 40 frames. The averaged phase indices (x-axis) are displayed together with the corre… view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of focus point Cmse location in relation to bi-ventricular segmentation at ED summarized across all dataset. For a better overview, the position per pathology is shown as a percentage per anatomical structure: LV (blue) - left ventricle, MYO (orange) - LV myocardium, RV (green) - right ventricle, ATRIA (red) - atria in 4CH and other in SAX. The cases which where not covered by the bi-ventricul… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

32 extracted references · 5 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , author Zhao, A

    author Balakrishnan, G. , author Zhao, A. , author Sabuncu, M.R. , author Guttag, J. , author Dalca, A.V. , year 2018 . title An unsupervised learning model for deformable medical image registration , in: booktitle Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 9252--9260

    2018

  3. [3]

    , author Moroni, D

    author Barcaro, U. , author Moroni, D. , author Salvetti, O. , year 2008 . title Automatic computation of left ventricle ejection fraction from dynamic ultrasound images . journal Pattern Recognition and Image Analysis volume 18 , pages 351--358

    2008

  4. [4]

    , author Lalande, A

    author Bernard, O. , author Lalande, A. , author Zotti, C. , author Cervenansky, F. , author Yang, X. , author Heng, P.A. , author Cetin, I. , author Lekadir, K. , author Camara, O. , author Gonzalez Ballester, M.A. , author Sanroma, G. , author Napel, S. , author Petersen, S. , author Tziritas, G. , author Grinias, E. , author Khened, M. , author Kollera...

    arXiv 2018

  5. [5]

    , author Gkontra, P

    author Campello, V.M. , author Gkontra, P. , author Izquierdo, C. , author Martín-Isla, C. , author Sojoudi, A. , author Full, P.M. , author Maier-Hein, K. , author Zhang, Y. , author He, Z. , author Ma, J. , author Parreño, M. , author Albiol, A. , author Kong, F. , author Shadden, S.C. , author Acero, J.C. , author Sundaresan, V. , author Saber, M. , au...

    arXiv 2021

  6. [6]

    , author Balakrishnan, G

    author Dalca, A.V. , author Balakrishnan, G. , author Guttag, J. , author Sabuncu, M.R. , year 2019 . title Unsupervised learning of probabilistic diffeomorphic registration for images and surfaces . journal Medical Image Analysis volume 57 , pages 226–236 . http://dx.doi.org/10.1016/j.media.2019.07.006, :10.1016/j.media.2019.07.006

    doi:10.1016/j.media.2019.07.006 2019

  7. [7]

    , author Behnam, H

    author Darvishi, S. , author Behnam, H. , author Pouladian, M. , author Samiei, N. , year 2013 . title Measuring left ventricular volumes in two-dimensional echocardiography image sequence using level-set method for automatic detection of end-diastole and end-systole frames . journal Research in Cardiovascular Medicine volume 2 , pages 39--45

    2013

  8. [8]

    , author Liao, Z

    author Dezaki, F.T. , author Liao, Z. , author Luong, C. , author Girgis, H. , author Dhungel, N. , author Abdi, A.H. , author Behnami, D. , author Gin, K. , author Rohling, R. , author Abolmaesumi, P. , et al., year 2018 . title Cardiac phase detection in echocardiograms with densely gated recurrent neural networks and global extrema loss . journal IEEE ...

    2018

  9. [9]

    , author stvik, A

    author Fiorito, A.M. , author stvik, A. , author Smistad, E. , author Leclerc, S. , author Bernard, O. , author Lovstakken, L. , year 2018 . title Detection of cardiac events in echocardiography using 3d convolutional recurrent neural networks , in: booktitle 2018 IEEE International Ultrasonics Symposium (IUS) , organization IEEE . pp. pages 1--4

    2018

  10. [10]

    , author Curran, K.M

    author Garcia-Cabrera, C. , author Curran, K.M. , author O'Connor, N.E. , author McGuinness, K. , year 2023 . title Cardiac magnetic resonance phase detection using neural networks , in: booktitle 2023 31st Irish Conference on Artificial Intelligence and Cognitive Science (AICS) , organization IEEE . pp. pages 1--4

    2023

  11. [11]

    , author Behnam, H

    author Gifani, P. , author Behnam, H. , author Shalbaf, A. , author Sani, Z. , year 2010 . title Automatic detection of end-diastole and end-systole from echocardiography images using manifold learning . journal Physiological Measurement volume 31 , pages 1091

    2010

  12. [12]

    , author Simonyan, K

    author Jaderberg, M. , author Simonyan, K. , author Zisserman, A. , author Kavukcuoglu, K. , year 2015 . title Spatial transformer networks , in: editor Cortes, C. , editor Lawrence, N. , editor Lee, D. , editor Sugiyama, M. , editor Garnett, R. (Eds.), booktitle Advances in Neural Information Processing Systems , publisher Curran Associates, Inc. . pp. p...

    Pith/arXiv arXiv doi:10.48550/arxiv.1506.02025 2015

  13. [13]

    , author Delouche, A

    author Kachenoura, N. , author Delouche, A. , author Herment, A. , author Frouin, F. , author Diebold, B. , year 2006 . title Automatic detection of end systole within a sequence of left ventricular echocardiographic images using autocorrelation and mitral valve motion detection , in: booktitle 2007 29th Annual International Conference of the IEEE Enginee...

    2006

  14. [14]

    , author Hussain, T

    author Koehler, S. , author Hussain, T. , author Hussain, H. , author Young, D. , author Sarikouch, S. , author Pickardt, T. , author Greil, G. , author Engelhardt, S. , year 2022 a. title Self-supervised motion descriptor for cardiac phase detection in 4d cmr based on discrete vector field estimations , in: booktitle International Workshop on Statistical...

    2022

  15. [15]

    , author Kuhm, J

    author Koehler, S. , author Kuhm, J. , author Huffaker, T. , author Young, D. , author Tandon, A. , author Andr \'e , F. , author Frey, N. , author Greil, G. , author Hussain, T. , author Engelhardt, S. , year 2025 . title Deep learning--based aligned strain from cine cardiac mri for detection of fibrotic myocardial tissue in patients with duchenne muscul...

    2025

  16. [16]

    , author Sharan, L

    author Koehler, S. , author Sharan, L. , author Kuhm, J. , author Ghanaat, A. , author Gordejeva, J. , author Simon, N.K. , author Grell, N.M. , author Andr \'e , F. , author Engelhardt, S. , year 2022 b. title Comparison of evaluation metrics for landmark detection in cmr images , in: booktitle Bildverarbeitung f \"u r die Medizin 2022: Proceedings, Germ...

    2022

  17. [17]

    , author Zhan, Y

    author Kong, B. , author Zhan, Y. , author Shin, M. , author Denny, T. , author Zhang, S. , year 2016 . title Recognizing end-diastole and end-systole frames via deep temporal regression network , in: booktitle Medical Image Computing and Computer-Assisted Intervention-MICCAI 2016: 19th International Conference, Athens, Greece, October 17-21, 2016, Procee...

    work page doi:10.1007/978-3-319-46726-9_31 2016

  18. [18]

    , author Delingette, H

    author Krebs, J. , author Delingette, H. , author Mailhé, B. , author Ayache, N. , author Mansi, T. , year 2019 . title Learning a probabilistic model for diffeomorphic registration . journal IEEE Transactions on Medical Imaging volume 38 , pages 2165--2176 . :10.1109/TMI.2019.2897112

    arXiv 2019

  19. [19]

    , author Mansi, T

    author Krebs, J. , author Mansi, T. , author Ayache, N. , author Delingette, H. , year 2020 . title Probabilistic motion modeling from medical image sequences: application to cardiac cine-mri , in: booktitle International Workshop on Statistical Atlases and Computational Models of the Heart , organization Springer . pp. pages 176--185 . :10.1007/978-3-030...

    work page doi:10.1007/978-3-030-39074-7_19 2020

  20. [20]

    , author Azarmehr, N

    author Lane, E.S. , author Azarmehr, N. , author Jevsikov, J. , author Howard, J.P. , author Shun-shin, M.J. , author Cole, G.D. , author Francis, D.P. , author Zolgharni, M. , year 2021 . title Multibeat echocardiographic phase detection using deep neural networks . journal Computers in Biology and Medicine volume 133 , pages 104373 . https://www.science...

    arXiv 2021

  21. [21]

    , author Lysyansky, P

    author Mada, R.O. , author Lysyansky, P. , author Daraban, A.M. , author Duchenne, J. , author Voigt, J.U. , year 2015 . title How to define end-diastole and end-systole? journal JACC: Cardiovascular Imaging volume 8 , pages 148--157 . https://www.jacc.org/doi/abs/10.1016/j.jcmg.2014.10.010, :10.1016/j.jcmg.2014.10.010, http://arxiv.org/abs/https://www.ja...

    work page doi:10.1016/j.jcmg.2014.10.010 2015

  22. [22]

    , author Campello, V.M

    author Mart \' n-Isla, C. , author Campello, V.M. , author Izquierdo, C. , author Kushibar, K. , author Sendra-Balcells, C. , author Gkontra, P. , author Sojoudi, A. , author Fulton, M.J. , author Arega, T.W. , author Punithakumar, K. , et al., year 2023 . title Deep learning segmentation of the right ventricle in cardiac mri: the m&ms challenge . journal...

    2023

  23. [23]

    , author Qin, C

    author Meng, Q. , author Qin, C. , author Bai, W. , author Liu, T. , author de Marvao, A. , author O’Regan, D.P. , author Rueckert, D. , year 2022 . title Mulvimotion: Shape-aware 3d myocardial motion tracking from multi-view cardiac mri . journal IEEE Transactions on Medical Imaging volume 41 , pages 1961--1974 . :10.1109/TMI.2022.3154599

    arXiv 2022

  24. [24]

    , author Bai, W

    author Qin, C. , author Bai, W. , author Schlemper, J. , author Petersen, S.E. , author Piechnik, S.K. , author Neubauer, S. , author Rueckert, D. , year 2018 . title Joint learning of motion estimation and segmentation for cardiac mr image sequences , in: booktitle Medical Image Computing and Computer Assisted Intervention--MICCAI 2018: 21st Internationa...

    doi:10.1007/978-3-030- 2018

  25. [25]

    , author Fischer, P

    author Ronneberger, O. , author Fischer, P. , author Brox, T. , year 2015 . title U-net: Convolutional networks for biomedical image segmentation , in: editor Navab, N. , editor Hornegger, J. , editor Wells, W.M. , editor Frangi, A.F. (Eds.), booktitle Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2015 , publisher Springer Internati...

    2015

  26. [26]

    , author Beerbaum, P

    author Sarikouch, S. , author Beerbaum, P. , year 2005 . title Follow up of post-repair tetralogy of fallot . https://clinicaltrials.gov/ct2/show/NCT00266188. en. note last accessed 03. June, 2024

    2005

  27. [27]

    , author Koerperich, H

    author Sarikouch, S. , author Koerperich, H. , author Dubowy, K.O. , author Boethig, D. , author Boettler, P. , author Mir, T.S. , author Peters, B. , author Kuehne, T. , author Beerbaum, P. , author for Congenital Heart Defects Investigators, G.C.N. , year 2011 . title Impact of gender and age on cardiovascular function late after repair of tetralogy of ...

    work page doi:10.1161/circ 2011

  28. [28]

    , author AlizadehSani, Z

    author Shalbaf, A. , author AlizadehSani, Z. , author Behnam, H. , year 2015 . title Echocardiography without electrocardiogram using nonlinear dimensionality reduction methods . journal Journal of Medical Ultrasonics volume 42 , pages 137--149

    2015

  29. [29]

    , year 2021

    author WHO, W.H.O. , year 2021 . title Cardiovascular diseases (cvds) . note https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds) (accessed: 11.02.2025)

    2021

  30. [30]

    , author Brahm, G

    author Xue, W. , author Brahm, G. , author Pandey, S. , author Leung, S. , author Li, S. , year 2018 . title Full left ventricle quantification via deep multitask relationships learning . journal Medical Image Analysis volume 43 , pages 54--65 . https://www.sciencedirect.com/science/article/pii/S1361841517301366, :https://doi.org/10.1016/j.media.2017.09.005

    work page doi:10.1016/j.media.2017.09.005 2018

  31. [31]

    , author He, Y

    author Yang, F. , author He, Y. , author Hussain, M. , author Xie, H. , author Lei, P. , year 2017 . title Convolutional neural network for the detection of end-diastole and end-systole frames in free-breathing cardiac magnetic resonance imaging . journal Computational and mathematical methods in medicine volume 2017 , pages 1640835

    2017

  32. [32]

    , author Negoita, M

    author Zolgharni, M. , author Negoita, M. , author Dhutia, N.M. , author Mielewczik, M. , author Manoharan, K. , author Sohaib, S.A. , author Finegold, J.A. , author Sacchi, S. , author Cole, G.D. , author Francis, D.P. , year 2017 . title 10 . journal Echocardiography volume 34 , pages 956--967

    2017