Evaluation of Spatial Attentive Deep Learning for Automatic Placental Segmentation on Longitudinal MRI

Magnetic Resonance Imaging-Based Radiomics of Axial and Sagittal Orientation in Pregnant Patients with Suspected Placenta Accreta Spectrum

RATIONALE AND OBJECTIVES

Placenta accreta spectrum (PAS) is associated with significant morbidity and mortality. Current radiomic analysis of PAS magnetic resonance (MR) images is often performed on a single imaging plane. However, depending on the chosen imaging plane, radiomic features extracted from the same patient may vary due to the differing orientations and anatomical contexts, potentially leading to inconsistent results. In this study, we applied region of interest (ROI)-based radiomic analysis on axial and sagittal MR images in pregnant patients at high risk for PAS. Our objective was to compare MR textural features extracted from these imaging planes and to correlate these findings with surgical outcomes, aiming to enhance the accuracy of PAS diagnosis and treatment planning.

MATERIALS AND METHODS

This is a retrospective review of MR images of pregnancies with prenatally suspected PAS. Volumetric placental, uterus, and internal os of the cervix regions of interest (ROI) were manually segmented on axial and sagittal MR images for each patient. Radiomic features were extracted following the image biomarker standardization initiative guideline. Agreement in features extracted from axial and sagittal images were assessed using Spearman's rank correlation coefficient.

RESULTS

Of the 101 pregnant patients that met the study inclusion criteria, 65 underwent cesarean hysterectomy for PAS. 77 percent of the radiomics features had strong Spearman rank correlations (>0.8) between axial and sagittal images, indicating that these imaging planes provide similar radiomics information. The diagnostic performance of features extracted from axial and sagittal planes was quantified under the receiver operating characteristics curve (AUC). We found that axial and sagittal planes have similar performance for the prediction of hysterectomy. Shape elongation, Placental Location within the Uterus (PLU), and heterogeneity features were significant predictors for hysterectomy regardless of the imaging plane.

CONCLUSION

Our study found that radiomics features extracted from axial and sagittal MR image plane in the same patient have excellent agreement and strong correlation. We identified several features present in both axial and sagittal images that were predictive in detecting PAS-suspected patient who required hysterectomy. These features may represent the underlying placental pathophysiology.

CSAM: A 2.5D Cross-Slice Attention Module for Anisotropic Volumetric Medical Image Segmentation

A large portion of volumetric medical data, especially magnetic resonance imaging (MRI) data, is anisotropic, as the through-plane resolution is typically much lower than the in-plane resolution. Both 3D and purely 2D deep learning-based segmentation methods are deficient in dealing with such volumetric data since the performance of 3D methods suffers when confronting anisotropic data, and 2D methods disregard crucial volumetric information. Insufficient work has been done on 2.5D methods, in which 2D convolution is mainly used in concert with volumetric information. These models focus on learning the relationship across slices, but typically have many parameters to train. We offer a Cross-Slice Attention Module (CSAM) with minimal trainable parameters, which captures information across all the slices in the volume by applying semantic, positional, and slice attention on deep feature maps at different scales. Our extensive experiments using different network architectures and tasks demonstrate the usefulness and generalizability of CSAM. Associated code is available at https://github.com/aL3x-O-o-Hung/CSAM.

Collaborative Learning for Annotation-Efficient Volumetric MR Image Segmentation

BACKGROUND

Deep learning has presented great potential in accurate MR image segmentation when enough labeled data are provided for network optimization. However, manually annotating three-dimensional (3D) MR images is tedious and time-consuming, requiring experts with rich domain knowledge and experience.

PURPOSE

To build a deep learning method exploring sparse annotations, namely only a single two-dimensional slice label for each 3D training MR image.

STUDY TYPE

Retrospective.

POPULATION

Three-dimensional MR images of 150 subjects from two publicly available datasets were included. Among them, 50 (1377 image slices) are for prostate segmentation. The other 100 (8800 image slices) are for left atrium segmentation. Five-fold cross-validation experiments were carried out utilizing the first dataset. For the second dataset, 80 subjects were used for training and 20 were used for testing.

FIELD STRENGTH/SEQUENCE

1.5 T and 3.0 T; axial T2-weighted and late gadolinium-enhanced, 3D respiratory navigated, inversion recovery prepared gradient echo pulse sequence.

ASSESSMENT

A collaborative learning method by integrating the strengths of semi-supervised and self-supervised learning schemes was developed. The method was trained using labeled central slices and unlabeled noncentral slices. Segmentation performance on testing set was reported quantitatively and qualitatively.

STATISTICAL TESTS

Quantitative evaluation metrics including boundary intersection-over-union (B-IoU), Dice similarity coefficient, average symmetric surface distance, and relative absolute volume difference were calculated. Paired t test was performed, and P < 0.05 was considered statistically significant.

RESULTS

Compared to fully supervised training with only the labeled central slice, mean teacher, uncertainty-aware mean teacher, deep co-training, interpolation consistency training (ICT), and ambiguity-consensus mean teacher, the proposed method achieved a substantial improvement in segmentation accuracy, increasing the mean B-IoU significantly by more than 10.0% for prostate segmentation (proposed method B-IoU: 70.3% ± 7.6% vs. ICT B-IoU: 60.3% ± 11.2%) and by more than 6.0% for left atrium segmentation (proposed method B-IoU: 66.1% ± 6.8% vs. ICT B-IoU: 60.1% ± 7.1%).

DATA CONCLUSIONS

A collaborative learning method trained using sparse annotations can segment prostate and left atrium with high accuracy.

LEVEL OF EVIDENCE

0 TECHNICAL EFFICACY: Stage 1.

The human placenta project: Funded studies, imaging technologies, and future directions

The placenta plays a critical role in fetal development. It serves as a multi-functional organ that protects and nurtures the fetus during pregnancy. However, despite its importance, the intricacies of placental structure and function in normal and diseased states have remained largely unexplored. Thus, in 2014, the National Institute of Child Health and Human Development launched the Human Placenta Project (HPP). As of May 2023, the HPP has awarded over $101 million in research funds, resulting in 41 funded studies and 459 publications. We conducted a comprehensive review of these studies and publications to identify areas of funded research, advances in those areas, limitations of current research, and continued areas of need. This paper will specifically review the funded studies by the HPP, followed by an in-depth discussion on advances and gaps within placental-focused imaging. We highlight the progress within magnetic reasonance imaging and ultrasound, including development of tools for the assessment of placental function and structure.