Concerns in using multi-detector computed tomography for diagnosing pulmonary embolism in daily practice. A cross-sectional analysis using expert opinion as reference standard

Seeking an optimal approach for Computer-aided Diagnosis of Pulmonary Embolism

Pulmonary Embolism (PE) represents a thrombus ("blood clot"), usually originating from a lower extremity vein, that travels to the blood vessels in the lung, causing vascular obstruction and in some patients death. This disorder is commonly diagnosed using Computed Tomography Pulmonary Angiography (CTPA). Deep learning holds great promise for the Computer-aided Diagnosis (CAD) of PE. However, numerous deep learning methods, such as Convolutional Neural Networks (CNN) and Transformer-based models, exist for a given task, causing great confusion regarding the development of CAD systems for PE. To address this confusion, we present a comprehensive analysis of competing deep learning methods applicable to PE diagnosis based on four datasets. First, we use the RSNA PE dataset, which includes (weak) slice-level and exam-level labels, for PE classification and diagnosis, respectively. At the slice level, we compare CNNs with the Vision Transformer (ViT) and the Swin Transformer. We also investigate the impact of self-supervised versus (fully) supervised ImageNet pre-training, and transfer learning over training models from scratch. Additionally, at the exam level, we compare sequence model learning with our proposed transformer-based architecture, Embedding-based ViT (E-ViT). For the second and third datasets, we utilize the CAD-PE Challenge Dataset and Ferdowsi University of Mashad's PE Dataset, where we convert (strong) clot-level masks into slice-level annotations to evaluate the optimal CNN model for slice-level PE classification. Finally, we use our in-house PE-CAD dataset, which contains (strong) clot-level masks. Here, we investigate the impact of our vessel-oriented image representations and self-supervised pre-training on PE false positive reduction at the clot level across image dimensions (2D, 2.5D, and 3D). Our experiments show that (1) transfer learning boosts performance despite differences between photographic images and CTPA scans; (2) self-supervised pre-training can surpass (fully) supervised pre-training; (3) transformer-based models demonstrate comparable performance but slower convergence compared with CNNs for slice-level PE classification; (4) model trained on the RSNA PE dataset demonstrates promising performance when tested on unseen datasets for slice-level PE classification; (5) our E-ViT framework excels in handling variable numbers of slices and outperforms sequence model learning for exam-level diagnosis; and (6) vessel-oriented image representation and self-supervised pre-training both enhance performance for PE false positive reduction across image dimensions. Our optimal approach surpasses state-of-the-art results on the RSNA PE dataset, enhancing AUC by 0.62% (slice-level) and 2.22% (exam-level). On our in-house PE-CAD dataset, 3D vessel-oriented images improve performance from 80.07% to 91.35%, a remarkable 11% gain. Codes are available at GitHub.com/JLiangLab/CAD_PE.

Sex differences in guideline-consistent diagnostic testing for acute pulmonary embolism among adult emergency department patients aged 18-49

BACKGROUND

Pulmonary embolism (PE) is a frequent diagnostic consideration in emergency department (ED) patients, yet diagnosis is challenging because symptoms of PE are nonspecific. Guidelines recommend the use of clinical decision tools to increase efficiency and avoid harms from overtesting, including D-dimer screening in patients not at high risk for PE. Women undergo testing for PE more often than men yet have a lower yield from testing. Our study objective was to determine whether patient sex influenced the odds of received guideline-consistent care.

METHODS

We performed a retrospective cohort study at two large U.S. academic EDs from January 1, 2016, to December 31, 2018. Nonpregnant patients aged 18-49 years were included if they presented with chest pain, shortness of breath, hemoptysis, or syncope and underwent testing for PE with D-dimer or imaging. Demographic and clinical data were exported from the electronic medical record (EMR). Pretest risk scores were calculated using manually abstracted EMR data. Diagnostic testing was then compared with recommended testing based on pretest risk. The primary outcome was receipt of guideline-consistent care, which required an elevated screening D-dimer prior to imaging in all non-high-risk patients.

RESULTS

We studied 1991 discrete patient encounters; 37% (735) of patients were male and 63% (1256) were female. Baseline characteristics, including revised Geneva scores, were similar between sexes. Female patients were more likely to receive guideline-consistent care (70% [874/1256] female vs. 63% [463/735] male, p < 0.01) and less likely to be diagnosed with PE (3.1% [39/1256] female vs. 5.3% [39/735] male, p < 0.05). The most common guideline deviation in both sexes was obtaining imaging without a screening D-dimer in a non-high-risk patient (75% [287/382] female vs. 75% [205/272] male).

CONCLUSIONS

In this cohort, females were more likely than males to receive care consistent with current guidelines and less likely to be diagnosed with PE.

Seeking an Optimal Approach for Computer-Aided Pulmonary Embolism Detection

Pulmonary embolism (PE) represents a thrombus ("blood clot"), usually originating from a lower extremity vein, that travels to the blood vessels in the lung, causing vascular obstruction and in some patients, death. This disorder is commonly diagnosed using CT pulmonary angiography (CTPA). Deep learning holds great promise for the computer-aided CTPA diagnosis (CAD) of PE. However, numerous competing methods for a given task in the deep learning literature exist, causing great confusion regarding the development of a CAD PE system. To address this confusion, we present a comprehensive analysis of competing deep learning methods applicable to PE diagnosis using CTPA at the both image and exam levels. At the image level, we compare convolutional neural networks (CNNs) with vision transformers, and contrast self-supervised learning (SSL) with supervised learning, followed by an evaluation of transfer learning compared with training from scratch. At the exam level, we focus on comparing conventional classification (CC) with multiple instance learning (MIL). Our extensive experiments consistently show: (1) transfer learning consistently boosts performance despite differences between natural images and CT scans, (2) transfer learning with SSL surpasses its supervised counterparts; (3) CNNs outperform vision transformers, which otherwise show satisfactory performance; and (4) CC is, surprisingly, superior to MIL. Compared with the state of the art, our optimal approach provides an AUC gain of 0.2% and 1.05% for image-level and exam-level, respectively.

Symptomatic subsegmental versus more central pulmonary embolism: Clinical outcomes during anticoagulation

BACKGROUND

The optimal therapy of patients with acute subsegmental pulmonary embolism (PE) is controversial.

METHODS

We used the RIETE (Registro Informatizado Enfermedad TromboEmbólica) database to compare the rate of symptomatic PE recurrences during anticoagulation in patients with subsegmental, segmental, or more central PEs.

RESULTS

Among 15 963 patients with a first episode of symptomatic PE, 834 (5.2%) had subsegmental PE, 3797 (24%) segmental, and 11 332 (71%) more central PE. Most patients in all subgroups received initial therapy with low-molecular-weight heparin, and then most switched to vitamin K antagonists. Median duration of therapy was 179, 185, and 204 days, respectively. During anticoagulation, 183 patients developed PE recurrences, 131 developed deep vein thrombosis (DVT), 543 bled, and 1718 died (fatal PE, 135). The rate of PE recurrences was twofold higher in patients with subsegmental PE than in those with segmental (hazard ratio [HR], 2.13; 95% confidence interval [CI], 1.16-3.85) or more central PE (HR, 1.89; 95% CI, 1.12-3.13). On multivariable analysis, patients with subsegmental PE had a higher risk for PE recurrences than those with central PE (adjusted HR, 1.75; 95% CI, 1.02-3.03). After stratifying patients with subsegmental PE according to ultrasound imaging in the lower limbs, the rate of PE recurrences was similar in patients with DVT, in patients without DVT, and in those with no ultrasound imaging.

CONCLUSIONS

Our study reveals that the risk for PE recurrences in patients with segmental PE is not lower than in those with more central PE, thus suggesting that the risk of PE recurrences is not influenced by the anatomic location of PE.

Antithrombotic Therapy for VTE Disease: CHEST Guideline and Expert Panel Report

BACKGROUND

We update recommendations on 12 topics that were in the 9th edition of these guidelines, and address 3 new topics.

METHODS

We generate strong (Grade 1) and weak (Grade 2) recommendations based on high- (Grade A), moderate- (Grade B), and low- (Grade C) quality evidence.

RESULTS

For VTE and no cancer, as long-term anticoagulant therapy, we suggest dabigatran (Grade 2B), rivaroxaban (Grade 2B), apixaban (Grade 2B), or edoxaban (Grade 2B) over vitamin K antagonist (VKA) therapy, and suggest VKA therapy over low-molecular-weight heparin (LMWH; Grade 2C). For VTE and cancer, we suggest LMWH over VKA (Grade 2B), dabigatran (Grade 2C), rivaroxaban (Grade 2C), apixaban (Grade 2C), or edoxaban (Grade 2C). We have not changed recommendations for who should stop anticoagulation at 3 months or receive extended therapy. For VTE treated with anticoagulants, we recommend against an inferior vena cava filter (Grade 1B). For DVT, we suggest not using compression stockings routinely to prevent PTS (Grade 2B). For subsegmental pulmonary embolism and no proximal DVT, we suggest clinical surveillance over anticoagulation with a low risk of recurrent VTE (Grade 2C), and anticoagulation over clinical surveillance with a high risk (Grade 2C). We suggest thrombolytic therapy for pulmonary embolism with hypotension (Grade 2B), and systemic therapy over catheter-directed thrombolysis (Grade 2C). For recurrent VTE on a non-LMWH anticoagulant, we suggest LMWH (Grade 2C); for recurrent VTE on LMWH, we suggest increasing the LMWH dose (Grade 2C).

CONCLUSIONS

Of 54 recommendations included in the 30 statements, 20 were strong and none was based on high-quality evidence, highlighting the need for further research.