Key result
A deep learning model trained with weakly annotated data achieved an ROC AUC of 0.94, with 93.5% specificity and 86.6% sensitivity in detecting pulmonary embolism from CT pulmonary angiograms.
Why the study?
To develop and evaluate a deep neural network model for automated detection of pulmonary embolism from CTPAs using only weakly labelled training data.
Does a deep learning model trained with weakly annotated data accurately detect pulmonary embolism from CTPAs?
Cohort (n=770)
No
Does a deep learning model trained with weakly annotated data accurately detect pulmonary embolism from CTPAs?
A deep learning model trained with a relatively small, weakly annotated dataset can achieve excellent performance in detecting pulmonary embolism from CTPAs.
Answers are grounded in this paper and cite their sources.
May support weakly supervised DL for PE detection on CTPA; leaves open prospective validation before clinical adoption.
BACKGROUND: The aim of this study was to develop and evaluate a deep neural network model in the automated detection of pulmonary embolism (PE) from computed tomography pulmonary angiograms (CTPAs) using only weakly labelled training data. METHODS: We developed a deep neural network model consisting of two parts: a convolutional neural network architecture called InceptionResNet V2 and a long-short term memory network to process whole CTPA stacks as sequences of slices. Two versions of the model were created using either chest X-rays (Model A) or natural images (Model B) as pre-training data. We retrospectively collected 600 CTPAs to use in training and validation and 200 CTPAs to use in testing. CTPAs were annotated only with binary labels on both stack- and slice-based levels. Performance of the models was evaluated with ROC and precision-recall curves, specificity, sensitivity, accuracy, as well as positive and negative predictive values. RESULTS: Both models performed well on both stack- and slice-based levels. On the stack-based level, Model A reached specificity and sensitivity of 93.5% and 86.6%, respectively, outperforming Model B slightly (specificity 90.7% and sensitivity 83.5%). However, the difference between their ROC AUC scores was not statistically significant (0.94 vs 0.91, p = 0.07). CONCLUSIONS: We show that a deep learning model trained with a relatively small, weakly annotated dataset can achieve excellent performance results in detecting PE from CTPAs.
No takes yet. Share an insight, caveat, or question.
Huhtanen et al. (2022) conducted a cohort in Pulmonary embolism (n=770). Deep learning model (InceptionResNet V2 and LSTM) vs. Manual radiologist interpretation (reference standard) was evaluated on Detection of pulmonary embolism on stack-based level (ROC AUC) (ROC AUC 0.94). A deep learning model trained with weakly annotated data achieved an ROC AUC of 0.94, with 93.5% specificity and 86.6% sensitivity in detecting pulmonary embolism from CT pulmonary angiograms.
Synapse has enriched 10 related papers. Consider them for comparative context: