# Convolutional Neural Network for Early Pulmonary Embolism Detection via Computed Tomography Pulmonary Angiography

arXiv: https://arxiv.org/abs/2204.03204 · HTML: https://mlsanity.com/paper/2204.03204/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 16 references that are in the map.
- First posted on arXiv 7 Apr 2022.

## Builds on (most cited first, 7 of 7)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Squeeze-and-Excitation Networks](https://mlsanity.com/paper/1709.01507/) (2017)
- [V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation](https://mlsanity.com/paper/1606.04797/) (2016)
- [Dilated Residual Networks](https://mlsanity.com/paper/1705.09914/) (2017)
- [FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation](https://mlsanity.com/paper/1903.11816/) (2019)
- [A Two-Stage Method for Text Line Detection in Historical Documents](https://mlsanity.com/paper/1802.03345/) (2018)
- [Calibrating the Adaptive Learning Rate to Improve Convergence of ADAM](https://mlsanity.com/paper/1908.00700/) (2019)

## Abstract (arXiv)

This study was conducted to develop a computer-aided detection (CAD) system for triaging patients with pulmonary embolism (PE). The purpose of the system was to reduce the death rate during the waiting period. Computed tomography pulmonary angiography (CTPA) is used for PE diagnosis. Because CTPA reports require a radiologist to review the case and suggest further management, this creates a waiting period during which patients may die. Our proposed CAD method was thus designed to triage patients with PE from those without PE. In contrast to related studies involving CAD systems that identify key PE lesion images to expedite PE diagnosis, our system comprises a novel classification-model ensemble for PE detection and a segmentation model for PE lesion labeling. The models were trained using data from National Cheng Kung University Hospital and open resources. The classification model yielded 0.73 for receiver operating characteristic curve (accuracy = 0.85), while the mean intersection over union was 0.689 for the segmentation model. The proposed CAD system can distinguish between patients with and without PE and automatically label PE lesions to expedite PE diagnosis