Anatomy-Guided Multitask Learning for MRI-Based Classification of Placenta Accreta Spectrum and its Subtypes

TL;DR

This work tackles the prenatal diagnosis of Placenta Accreta Spectrum (PAS) and its subtypes using MRI by introducing a one-stage multiclass CNN with two branches and multitask learning. The model combines a residual-block classification backbone with a ROI-guided segmentation decoder, optimizing a joint loss $L = L_{cls} + \lambda L_{seg}$ (with $\lambda=1$) to perform four-class classification ($C=4$) among non-PAS, PA, PI, and PP while also predicting ROI masks. On a real clinical dataset of 4,140 MRI slices from 414–416 patients, the approach achieves state-of-the-art macro-AUC ($0.8015$) and demonstrates the benefit of incorporating anatomical ROI guidance and multitask learning. These results suggest a practical, efficient framework for prenatal PAS subtype diagnosis with potential clinical impact for risk stratification and management.

Abstract

Placenta Accreta Spectrum Disorders (PAS) pose significant risks during pregnancy, frequently leading to postpartum hemorrhage during cesarean deliveries and other severe clinical complications, with bleeding severity correlating to the degree of placental invasion. Consequently, accurate prenatal diagnosis of PAS and its subtypes-placenta accreta (PA), placenta increta (PI), and placenta percreta (PP)-is crucial. However, existing guidelines and methodologies predominantly focus on the presence of PAS, with limited research addressing subtype recognition. Additionally, previous multi-class diagnostic efforts have primarily relied on inefficient two-stage cascaded binary classification tasks. In this study, we propose a novel convolutional neural network (CNN) architecture designed for efficient one-stage multiclass diagnosis of PAS and its subtypes, based on 4,140 magnetic resonance imaging (MRI) slices. Our model features two branches: the main classification branch utilizes a residual block architecture comprising multiple residual blocks, while the second branch integrates anatomical features of the uteroplacental area and the adjacent uterine serous layer to enhance the model's attention during classification. Furthermore, we implement a multitask learning strategy to leverage both branches effectively. Experiments conducted on a real clinical dataset demonstrate that our model achieves state-of-the-art performance.

Figures (1)

• Figure 1: An overview of the proposed model framework. The backbone of our model is the classification branch, located on the left of the structure; it contains four residual blocks and also works as the segmentation branch's encoder. The line in purple means the skip connection. The ball in the green means the add operator.