Deep learning for pulmonary embolism detection on computed tomography pulmonary angiogram: a systematic review and meta-analysis

Abstract

Computed tomographic pulmonary angiography (CTPA) is the gold standard for pulmonary embolism (PE) diagnosis. However, this diagnosis is susceptible to misdiagnosis. In this study, we aimed to perform a systematic review of current literature applying deep learning for the diagnosis of PE on CTPA. MEDLINE/PUBMED were searched for studies that reported on the accuracy of deep learning algorithms for PE on CTPA. The risk of bias was evaluated using the QUADAS-2 tool. Pooled sensitivity and specificity were calculated. Summary receiver operating characteristic curves were plotted. Seven studies met our inclusion criteria. A total of 36,847 CTPA studies were analyzed. All studies were retrospective. Five studies provided enough data to calculate summary estimates. The pooled sensitivity and specificity for PE detection were 0.88 (95% CI 0.803-0.927) and 0.86 (95% CI 0.756-0.924), respectively. Most studies had a high risk of bias. Our study suggests that deep learning models can detect PE on CTPA with satisfactory sensitivity and an acceptable number of false positive cases. Yet, these are only preliminary retrospective works, indicating the need for future research to determine the clinical impact of automated PE detection on patient care. Deep learning models are gradually being implemented in hospital systems, and it is important to understand the strengths and limitations of these algorithms.

Figure 1

Artificial intelligence (AI) is an umbrella of terms encompassing machine learning and deep learning.

Figure 2

Comparison between artificial and biologic neural networks. Neural networks are comprised of multiple interconnected layers. Data is fed to the network, and an output is produced. By comparing the network’s output to the desired true label, an error can be estimated. Based on the error, the algorithm optimizes connections between the layers. The connections between the neurons are termed “weights”. Ultimately, a tuned network is achieved.

Figure 3

The architecture of Convolutional Neural Network (CNN). CNNs are networks specifically designed to process images. Many small filters compose each CNN layer. A filter is a small matrix of weights that is repeatedly applied to the image pixels. By sharing the filter across the image, repeating patterns are recognized. CNNs are ideal for image analysis since images are composed of repeating patterns. The shallow layers of the CNN recognize low-level patterns. The deeper layers gain a high-level understanding of the image.

Figure 4

Main computer vision tasks: classification, detection, and segmentation.

Figure 5

Flow diagram of the search and inclusion process.

Figure 6

(A) Sensitivity and Specificity of included studies (B) Bivariate summary ROC curves for the detection of pulmonary embolism on CTPA using deep learning.