Rethinking Pulmonary Embolism Segmentation: A Study of Current Approaches and Challenges with an Open Weight Model

2025-09-22

AI Summary

Pulmonary embolism (PE) segmentation faces challenges in detecting distal small emboli and suffers from poor model generalizability. Method: We constructed a high-quality dataset of 490 CTPA scans and systematically benchmarked nine 3D segmentation architectures. We identified detrimental negative transfer from classification pretraining—indicating feature heterogeneity between PE classification and segmentation—and empirically demonstrated the superiority of CNN-based models (especially 3D U-Net with ResNet encoders) over ViT-based architectures. A unified ablation framework was proposed to isolate key design factors, revealing consistent scale-dependent performance patterns. Results: The optimal model achieved a Dice score of 0.713 on the internal test set, detecting 181 emboli (FP = 49, FN = 28), and demonstrated strong generalizability on public benchmarks. Core contributions include: (1) the first large-scale, standardized PE segmentation benchmark; (2) empirical evidence of pretraining-induced negative transfer; and (3) a systematic analysis of the distal embolus detection bottleneck.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

In this study, we curated a densely annotated in-house dataset comprising 490 CTPA scans. Using this dataset, we systematically evaluated nine widely used segmentation architectures from both the CNN and Vision Transformer (ViT) families, initialized with either pretrained or random weights, under a unified testing framework as a performance audit. Our study leads to several important observations: (1) 3D U-Net with a ResNet encoder remains a highly effective architecture for PE segmentation; (2) 3D models are particularly well-suited to this task given the morphological characteristics of emboli; (3) CNN-based models generally yield superior performance compared to their ViT-based counterparts in PE segmentation; (4) classification-based pretraining, even on large PE datasets, can adversely impact segmentation performance compared to training from scratch, suggesting that PE classification and segmentation may rely on different sets of discriminative features; (5) different model architectures show a highly consistent pattern of segmentation performance when trained on the same data; and (6) while central and large emboli can be segmented with satisfactory accuracy, distal emboli remain challenging due to both task complexity and the scarcity of high-quality datasets. Besides these findings, our best-performing model achieves a mean Dice score of 0.7131 for segmentation. It detects 181 emboli with 49 false positives and 28 false negatives from 60 in-house testing scans. Its generalizability is further validated on public datasets.

Problem

• Evaluating segmentation architectures for pulmonary embolism detection
• Comparing CNN and Vision Transformer performance on PE segmentation
• Addressing challenges in segmenting distal emboli from CTPA scans

Innovation

• Evaluated nine segmentation architectures systematically
• Found 3D U-Net with ResNet encoder most effective
• Showed classification pretraining harms segmentation performance