Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Placenta Accreta Spectrum Detection Using an MRI-based Hybrid CNN-Transformer Model

Sumaiya Ali, Areej Alhothali, Ohoud Alzamzami, Sameera Albasri, Ahmed Abduljabbar, Muhammad Alwazzan

TL;DR

This study tackles PAS diagnosis from MRI by developing a fully end-to-end 3D hybrid CNN–Transformer model (DenseNet121–ViT) that processes volumetric scans to capture local textures and global spatial context. In a retrospective cohort of 1,133 MRI volumes, the proposed DenseNet121–ViT outperformed six other 3D architectures, achieving a five-run average accuracy of 84.3% and AUC of 0.842 on independent test data, with a robust, clinically relevant performance profile. The work demonstrates the potential of hybrid 3D architectures to improve diagnostic consistency and timeliness in PAS, while acknowledging the need for multi-center validation and interpretability to enable real-world adoption.

Abstract

Placenta Accreta Spectrum (PAS) is a serious obstetric condition that can be challenging to diagnose with Magnetic Resonance Imaging (MRI) due to variability in radiologists' interpretations. To overcome this challenge, a hybrid 3D deep learning model for automated PAS detection from volumetric MRI scans is proposed in this study. The model integrates a 3D DenseNet121 to capture local features and a 3D Vision Transformer (ViT) to model global spatial context. It was developed and evaluated on a retrospective dataset of 1,133 MRI volumes. Multiple 3D deep learning architectures were also evaluated for comparison. On an independent test set, the DenseNet121-ViT model achieved the highest performance with a five-run average accuracy of 84.3%. These results highlight the strength of hybrid CNN-Transformer models as a computer-aided diagnosis tool. The model's performance demonstrates a clear potential to assist radiologists by providing a robust decision support to improve diagnostic consistency across interpretations, and ultimately enhance the accuracy and timeliness of PAS diagnosis.

Placenta Accreta Spectrum Detection Using an MRI-based Hybrid CNN-Transformer Model

TL;DR

This study tackles PAS diagnosis from MRI by developing a fully end-to-end 3D hybrid CNN–Transformer model (DenseNet121–ViT) that processes volumetric scans to capture local textures and global spatial context. In a retrospective cohort of 1,133 MRI volumes, the proposed DenseNet121–ViT outperformed six other 3D architectures, achieving a five-run average accuracy of 84.3% and AUC of 0.842 on independent test data, with a robust, clinically relevant performance profile. The work demonstrates the potential of hybrid 3D architectures to improve diagnostic consistency and timeliness in PAS, while acknowledging the need for multi-center validation and interpretability to enable real-world adoption.

Abstract

Placenta Accreta Spectrum (PAS) is a serious obstetric condition that can be challenging to diagnose with Magnetic Resonance Imaging (MRI) due to variability in radiologists' interpretations. To overcome this challenge, a hybrid 3D deep learning model for automated PAS detection from volumetric MRI scans is proposed in this study. The model integrates a 3D DenseNet121 to capture local features and a 3D Vision Transformer (ViT) to model global spatial context. It was developed and evaluated on a retrospective dataset of 1,133 MRI volumes. Multiple 3D deep learning architectures were also evaluated for comparison. On an independent test set, the DenseNet121-ViT model achieved the highest performance with a five-run average accuracy of 84.3%. These results highlight the strength of hybrid CNN-Transformer models as a computer-aided diagnosis tool. The model's performance demonstrates a clear potential to assist radiologists by providing a robust decision support to improve diagnostic consistency across interpretations, and ultimately enhance the accuracy and timeliness of PAS diagnosis.

Paper Structure

This paper contains 13 sections, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Radiomic Approach
  4. Deep Learning
  5. Hybrid Learning
  6. MATERIAL AND METHODS
  7. Dataset
  8. Deep Learning Models
  9. Experimental Setup
  10. Data Preparation
  11. Model Training and Optimization
  12. Results and Discussion
  13. Conclusion

Figures (7)

  • Figure 1: MRI signs of PAS: (A) Intraplacental bands, (B) Myometrial border interruption, and (C) Abnormal vascularity. b7
  • Figure 2: Example MRI slices from the dataset: (A) Normal case, and (B) PAS case.
  • Figure 3: Architecture of the proposed hybrid 3D DenseNet121-ViT model.
  • Figure 4: Workflow of the proposed 3D MRI-based PAS classification pipeline.
  • Figure 5: Preprocessing pipeline for standardizing MRI scans.
  • ...and 2 more figures