MétaCan
Menu
Back to cohort
Record W4223914566 · doi:10.1002/mp.15670

Uncertainty‐guided symmetric multilevel supervision network for 3D left atrium segmentation in late gadolinium‐enhanced MRI

2022· article· en· W4223914566 on OpenAlexaff
Yashu Liu, Wei Wang, Gongning Luo, Kuanquan Wang, Dong Liang, Shuo Li

Bibliographic record

VenueMedical Physics · 2022
Typearticle
Languageen
FieldComputer Science
TopicMedical Image Segmentation Techniques
Canadian institutionsWestern University
FundersNatural Science Foundation of Shandong ProvinceChina Postdoctoral Science FoundationNational Natural Science Foundation of ChinaHeilongjiang Postdoctoral Science Foundation
KeywordsSegmentationJaccard indexComputer scienceArtificial intelligenceHausdorff distanceMagnetic resonance imagingConvolutional neural networkPattern recognition (psychology)Image segmentationMedicineRadiology

Abstract

fetched live from OpenAlex

PURPOSE: Atrial fibrillation (AF) is a common arrhythmia and requires volumetric imaging to guide the therapy procedure. Late gadolinium-enhanced magnetic resonance imaging (LGE MRI) is an efficient noninvasive technology for imaging the diseased heart. Three-dimensional segmentation of the left atrium (LA) in LGE MRI is a fundamental step for guiding the therapy of patients with AF. However, the low contrast and fuzzy surface of the LA in LGE MRI make accurate and objective LA segmentation a challenge. The purpose of this study is to propose an automatic and efficient LA segmentation model based on a convolutional neural network to obtain a more accurate predicted surface and improve the LA segmentation results. METHODS: In this study, we proposed an uncertainty-guided symmetric multilevel supervision (SML) network for 3D LA segmentation in LGE MRI. First, we constructed an SML structure to combine the corresponding features from the encoding and decoding stages to learn the multiscale representation of LA. Second, we formulated the discrepancy of predictions of our model as model uncertainty. Then we proposed an uncertainty-guided objective function to further increase the segmentation accuracy on the surface. RESULTS: We evaluated our proposed model on the public LA segmentation database using four universal metrics. The proposed model achieved Hausdorff Distance (HD) of 11.68 mm, average symmetric surface distance of 0.92 mm, Dice score of 0.92, and Jaccard of 0.85. Compared with state-of-the-art models, our model achieved the best HD that is sensitive to surface accuracy. For the other three metrics, our model also achieved better or comparable performance. CONCLUSIONS: We proposed an efficient automatic LA segmentation model that consisted of an SML structure and an uncertainty-guided objective function. Compared to other models, we designed an additional supervision branch in the encoding stage to learn more detailed representations of LA while learning global context information through the multilevel structure of each supervision branch. To address the fuzzy surface challenge of LA segmentation in LGE MRI, we leveraged the model uncertainty to enhance the distinguishing ability of the model on the surface, thereby the predicted accuracy of the LA surface can be further increased. We conducted extensive ablation and comparative experiments with state-of-the-art models. The experiment results demonstrated that our proposed model could handle the complex structure of LA and had superior advantages in improving the segmentation performance on the surface.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.002
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: none
Teacher disagreement score0.954
Threshold uncertainty score0.841

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0020.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.001
Science and technology studies0.0000.000
Scholarly communication0.0000.000
Open science0.0010.001
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.028
GPT teacher head0.317
Teacher spread0.288 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one teacher head, not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreMethods

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations8
Published2022
Admission routes1
Has abstractyes

Explore more

Same venueMedical PhysicsSame topicMedical Image Segmentation TechniquesFrench-language works237,207