From Point Estimates to Distributions: GMM Pooling for MIL in Preterm Birth Prediction

2026-06-22

AI Summary

This work addresses the limitation in preterm birth prediction of relying solely on a single ultrasound image while neglecting longitudinal scan information. To capture intra-patient variability, the authors formulate the problem as a multiple instance learning (MIL) task, modeling each patient’s set of transvaginal ultrasound images as a bag of instances. They propose a novel Gaussian Mixture Model (GMM)-based pooling approach that explicitly characterizes the distribution of image features within each bag, generating fixed-length representations that go beyond conventional MIL methods which collapse bags into point estimates. Evaluated on a private preterm birth dataset, the method improves the PR-AUC from 0.44 to 0.56. Furthermore, it achieves state-of-the-art performance on a public lymph node metastasis benchmark, reporting an F1-score of 0.91, ROC-AUC of 0.89, and regression MAE of 0.18.

Abstract

Preterm birth (PTB) prediction can enable targeted surveillance and timely intervention, yet most ultrasound-based models use a single selected transvaginal ultrasound (TVUS) frame per patient despite routine exams acquiring multiple cervical images. We formulate PTB prediction as a multiple instance learning (MIL) problem, representing each patient as a variable-sized bag of TVUS images with a single outcome label. To move beyond standard MIL aggregators that collapse a bag into a point estimate, we propose a Gaussian Mixture Model (GMM) pooling, which summarizes all images in a bag into a fixed-length representation by modeling their feature distribution. This design captures intra-patient variability. We evaluate the method on a private clinical cohort and on a public lymph node metastasis benchmark. For PTB prediction, GMM pooling improves over the instance-based model PR-AUC from 0.44 to 0.56. On the lymph node benchmark, it achieves state-of-the-art performance with 0.91 F1-score and 0.89 ROC-AUC for classification and 0.18 MAE for regression. The code is publicly available at https://github.com/HussainAlasmawi/GMM_Pooling.

Problem

• preterm birth prediction
• multiple instance learning
• ultrasound imaging
• intra-patient variability
• cervical images

Innovation

• Gaussian Mixture Model
• Multiple Instance Learning
• Distribution-based Pooling
• Preterm Birth Prediction
• Intra-patient Variability