CoChat

Discriminative Learning for Automatic Staging of Placental Maturity via Multi-layer Fisher Vector

Open access

The DOI matches its Crossref record; 42 of the 42 references it cites were found in a registry or an index; 4 of the 11 sources its claims cite were checked for support.

Abstract

This paper proposes a method to automatically stage placental maturity from B-mode ultrasound images based on dense sampling and multi-layer Fisher vector encoding.

Overview

CoChat Summary

## TL;DR

This paper proposes an automatic placental maturity staging method from B-mode ultrasound images using dense sampling and Fisher vector encoding, with a multi-layer strategy to incorporate spatial information. On 443 placental images, the dense FV approach achieved an AUC of 96.77%, sensitivity of 98.04%, and specificity of 93.75%.

## Key findings

  • The proposed dense FV method achieved AUC = 96.77%, sensitivity = 98.04%, and specificity = 93.75% for placental maturity staging.
  • Compared with other encoding methods on DSIFT features, FV outperformed VLAD (AUC 95.3%) and BoVW (AUC 90.62%).
  • The paper reports that dense features outperformed traditional sparse features for placental maturity staging.
  • Among tested feature descriptors, DAISY achieved the best staging performance; the paper also states that W = 400 gave a good trade-off between performance and efficacy.

## Methods

  • The study used 443 placental B-mode ultrasound images from Shenzhen Maternal and Child Health Hospital: 187 stage 0, 135 stage 1, 85 stage 2, and 36 stage 3; gestational age ranged from 18 to 40 weeks.
  • The pipeline used a BoF framework with dense sampling on a regular grid, feature descriptors including SIFT, raw intensity, LIOP, DAISY, and SIFT+intensity, and Fisher vector encoding; a multi-layer / spatial pyramid strategy was used to add spatial information.
  • Performance was evaluated with 10-fold cross-validation, repeated at least 10 times, using SVM classifiers and metrics including AUC, sensitivity, specificity, mAP, and ROC curves.

## Limitations

  • The paper states that none of the existing methods had been applied for clinical practice at the time of writing, indicating limited clinical validation.
  • The text makes evident that the dataset is small and class-imbalanced across the four stages, especially stage 3 with 36 images (evident from the sample distribution).
  • The paper suggests future work is needed, including hierarchical fusion of dense and sparse features, blood flow information, classifier fusion, and deep learning, which implies the current method is not yet fully optimized (stated in the paper).

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/srep12818

Cites

42 of 42 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Image samples of 4 stages (from left to right, stage 0 to stage 3 image).
Figure 1 Image samples of 4 stages (from left to right, stage 0 to stage 3 image).CC BY · PubMed Central
Figure 2 Different feature detector methods. (a) Dense sampling; (b) MultiscaleHarris; (c) MultiscaleHessian; (d) HarrisLaplace; (e) DoG; (f) Hessian.
Figure 2 Different feature detector methods. (a) Dense sampling; (b) MultiscaleHarris; (c) MultiscaleHessian; (d) HarrisLaplace; (e) DoG; (f) Hessian.CC BY · PubMed Central
Figure 3 Staging results of different feature detectors in terms of Hessian, HarrisLaplace, HessianLaplace, MultiscaleHessian, MultiscaleHarris, DoG and dense sampling method. (a) mAP, sensitivity, specificity results of different feature detectors; (b) RoC curves of different feature detectors.
Figure 3 Staging results of different feature detectors in terms of Hessian, HarrisLaplace, HessianLaplace, MultiscaleHessian, MultiscaleHarris, DoG and dense sampling method. (a) mAP, sensitivity, specificity results of different feature detectors; (b) RoC curves of different feature detectors.CC BY · PubMed Central
Figure 4 Staging results of different feature descriptors using dense sampling method. (a) mAP, sensitivity, specificity results of different feature descriptors; (b) RoC curves of different feature descriptors.
Figure 4 Staging results of different feature descriptors using dense sampling method. (a) mAP, sensitivity, specificity results of different feature descriptors; (b) RoC curves of different feature descriptors.CC BY · PubMed Central
Figure 5 Staging results of SIFT descriptor using different vocabulary sizes in terms of W = 200, W = 400, W = 600, W = 800, and W = 1000. (a) mAP, sensitivity, specificity results of different vocabulary sizes; (b) RoC curves of different vocabulary sizes.
Figure 5 Staging results of SIFT descriptor using different vocabulary sizes in terms of W = 200, W = 400, W = 600, W = 800, and W = 1000. (a) mAP, sensitivity, specificity results of different vocabulary sizes; (b) RoC curves of different vocabulary sizes.CC BY · PubMed Central
Figure 6 Staging results of feature encoding methods using SIFT descriptors. (a) mAP, sensitivity, specificity results of different feature encoding methods; (b) RoC curves of different feature encoding methods.
Figure 6 Staging results of feature encoding methods using SIFT descriptors. (a) mAP, sensitivity, specificity results of different feature encoding methods; (b) RoC curves of different feature encoding methods.CC BY · PubMed Central
Figure 7 Query and top seven retrieved medical images. The first image is the query image, whereas the rest are the retrieved placental image based on the similarity scores (ranked from left to right, top to bottom). (a) Stage 1, (b) Stage 2, (c) Stage 3, (d) Stage 4.
Figure 7 Query and top seven retrieved medical images. The first image is the query image, whereas the rest are the retrieved placental image based on the similarity scores (ranked from left to right, top to bottom). (a) Stage 1, (b) Stage 2, (c) Stage 3, (d) Stage 4.CC BY · PubMed Central
Figure 8 System overview of the automatic placental maturity staging method. The input images are first pre-processed (i.e., noise reduction), and then the features are extracted based on the dense sampling on the pre-processed placental images. FV is investigated to encode the extracted features and transform them to high-level representation in terms of histogram of occurrence.
Figure 8 System overview of the automatic placental maturity staging method. The input images are first pre-processed (i.e., noise reduction), and then the features are extracted based on the dense sampling on the pre-processed placental images. FV is investigated to encode the extracted features and transform them to high-level representation in terms of histogram of occurrence.CC BY · PubMed Central
Figure 9 The procedure of feature vector formation. The input placental image is first partitioned into patches, and each patch is represented by the patch descriptor. GMM is applied to generate k Gaussians based on the assumption of a diagonal covariance matrix. The extracted feature is encoded by FV into histogram of occurrence as feature vector.
Figure 9 The procedure of feature vector formation. The input placental image is first partitioned into patches, and each patch is represented by the patch descriptor. GMM is applied to generate k Gaussians based on the assumption of a diagonal covariance matrix. The extracted feature is encoded by FV into histogram of occurrence as feature vector.CC BY · PubMed Central
Figure 10 Illustration of dense sampling with SIFT descriptor.
Figure 10 Illustration of dense sampling with SIFT descriptor.CC BY · PubMed Central
Figure 11 Layout of the LIOP descriptor, where shaded area is input patch (square area), white area is circular measurement region, and blue area is local neighborhood of a point.
Figure 11 Layout of the LIOP descriptor, where shaded area is input patch (square area), white area is circular measurement region, and blue area is local neighborhood of a point.CC BY · PubMed Central

Verification

Claims the paper makes

2 Valid · 2 Not valid · 5 Not checked · 45 Own results and interpretations

25 more claims

Create a free account to see all 54 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 13
  • It may state two separate claims as one. 14

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Show all 42 references

Research works printed but not linked: 1

  1. Chen, N., Jiang, F. & Luo, F .-c. Research progress of ultrasound evaluating fetal lung maturity. Anhui Med and Pharmaceutical Journal. 2, 50 (2011).

Cited by

See all 28

Metadata

Authors
Lei, B., Yao, Y., Chen, S., Li, S., Li, W., Ni, D., & Wang, T.
Affiliations
  • Shenzhen University (CN)
  • Shenzhen Maternity and Child Healthcare Hospital (CN)
Corresponding author
  • Ni, Dong
  • Wang, Tianfu
Venue
Scientific Reports
Published
2015-07-31
Volume
5
Issue
1
Pages
12818-12818
ISSN
2045-2322 (ISSN-L)
OpenAlex
W2418255636
Cited by
28
References
42
Citations by year
2026: 2 · 2025: 2 · 2024: 1 · 2023: 2 · 2022: 1 · 2021: 2 · 2020: 2 · 2018: 3 · 2017: 7 · 2016: 5 · 2015: 1
FWCI
3.84
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • National Natural Science Foundation of China (81270707, 61402296, 61101026, 61427806, 61372006)
  • Natural Science Foundation of Guangdong Province (S2013040014448)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image Retrieval and Classification Techniques
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Radiomics and Machine Learning in Medical Imaging
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Medical Image Segmentation Techniques
Keywords
placental maturity, fetal biometry, B-mode ultrasound, Fisher vectors, Gaussian mixture model, feature clustering, discriminative learning, ultrasound image analysis
MeSH terms
Female, Humans, Image Interpretation, Computer-Assisted, Placenta, Pregnancy, Cluster Analysis, Ultrasonography, Prenatal, Support Vector Machine

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Discriminative Learning for Automatic Staging of Placental Maturity via Multi-layer Fisher Vector. https://app.cochat.ai/papers/10.1038/srep12818
The paper
Lei, B., Yao, Y., Chen, S., Li, S., Li, W., Ni, D., & Wang, T. (2015). Discriminative Learning for Automatic Staging of Placental Maturity via Multi-layer Fisher Vector. Scientific Reports. https://doi.org/10.1038/srep12818

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.