skip to main content

Title: Attention Based CNN-LSTM Network for Pulmonary Embolism Prediction on Chest Computed Tomography Pulmonary Angiograms
Award ID(s):
Author(s) / Creator(s):
; ; ; ; ; ; ;
Date Published:
Journal Name:
International Conference on Medical Image Computing and Computer-Assisted Intervention.
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Pulmonary vascular distensibility (α) is a marker of the ability of the pulmonary vasculature to dilate in response to increases in cardiac output, which protects the right ventricle from excessive increases in afterload. α measured with exercise predicts clinical outcomes in pulmonary hypertension (PH) and heart failure. In this study, we aim to determine if α measured with a passive leg raise (PLR) maneuver is comparable to α with exercise. Invasive cardiopulmonary exercise testing (iCPET) was performed with hemodynamics recorded at three stages: rest, PLR and peak exercise. Four hemodynamic phenotypes were identified (2019 ECS guidelines): pulmonary arterial hypertension (PAH) (n = 10), isolated post‐capillary (Ipc‐PH) (n = 18), combined pre‐/post‐capillary PH (Cpc‐PH) (n = 15), and Control (no significant PH at rest and exercise) (n = 7). Measurements of mean pulmonary artery pressure, pulmonary artery wedge pressure, and cardiac output at each stage were used to calculate α. There was no statistical difference between α‐exercise and α‐PLR (0.87 ± 0.68 and 0.78 ± 0.47% per mmHg, respectively). The peak exercise‐ and PLR‐based calculations of α among the four hemodynamic groups were: Ipc‐PH = Ex: 0.94 ± 0.30, PLR: 1.00 ± 0.27% per mmHg; Cpc‐PH = Ex: 0.51 ± 0.15, PLR: 0.47 ± 0.18% per mmHg; PAH = Ex: 0.39 ± 0.23, PLR: 0.34 ± 0.18% per mmHg; and the Control group: Ex: 2.13 ± 0.91, PLR: 1.45 ± 0.49% per mmHg. Patients withα ≥ 0.7% per mmHg had reduced cardiovascular death and hospital admissions at 12‐month follow‐up. In conclusion, α‐PLR is feasible and may be equally predictive of clinical outcomes as α‐exercise in patients who are unable to exercise or in programs lacking iCPET facilities.

    more » « less
  2. Key points

    Imaging techniques such as contrast echocardiography suggest that anatomical intra‐pulmonary arteriovenous anastomoses (IPAVAs) are present at rest and are recruited to a greater extent in conditions such as exercise. IPAVAs have the potential to act as a shunt, although gas exchange methods have not demonstrated significant shunt in the normal lung.

    To evaluate this discrepancy, we compared anatomical shunt with 25‐µm microspheres to contrast echocardiography, and gas exchange shunt measured by the multiple inert gas elimination technique (MIGET).

    Intra‐pulmonary shunt measured by 25‐µm microspheres was not significantly different from gas exchange shunt determined by MIGET, suggesting that MIGET does not underestimate the gas exchange consequences of anatomical shunt.

    A positive agitated saline contrast echocardiography score was associated with anatomical shunt measured by microspheres. Agitated saline contrast echocardiography had high sensitivity but low specificity to detect a ≥1% anatomical shunt, frequently detecting small shunts inconsequential for gas exchange.


    The echocardiographic visualization of transpulmonary agitated saline microbubbles suggests that anatomical intra‐pulmonary arteriovenous anastomoses are recruited during exercise, in hypoxia, and when cardiac output is increased pharmacologically. However, the multiple inert gas elimination technique (MIGET) shows insignificant right‐to‐left gas exchange shunt in normal humans and canines. To evaluate this discrepancy, we measured anatomical shunt with 25‐µm microspheres and compared the results to contrast echocardiography and MIGET‐determined gas exchange shunt in nine anaesthetized, ventilated canines. Data were acquired under the following conditions: (1) at baseline, (2) 2 µg kg−1 min−1i.v.dopamine, (3) 10 µg kg−1 min−1i.v.dobutamine, and (4) following creation of an intra‐atrial shunt (in four animals). Right to left anatomical shunt was quantified by the number of 25‐µm microspheres recovered in systemic arterial blood. Ventilation–perfusion mismatch and gas exchange shunt were quantified by MIGET and cardiac output by direct Fick. Left ventricular contrast scores were assessed by agitated saline bubble counts, and separately by appearance of 25‐µm microspheres. Across all conditions, anatomical shunt measured by 25‐µm microspheres was not different from gas exchange shunt measured by MIGET (microspheres: 2.3 ± 7.4%; MIGET: 2.6 ± 6.1%,P = 0.64). Saline contrast bubble score was associated with microsphere shunt (ρ = 0.60,P < 0.001). Agitated saline contrast score had high sensitivity (100%) to detect a ≥1% shunt, but low specificity (22–48%). Gas exchange shunt by MIGET does not underestimate anatomical shunt measured using 25‐µm microspheres. Contrast echocardiography is extremely sensitive, but not specific, often detecting small anatomical shunts which are inconsequential for gas exchange.

    more » « less
  3. Abstract Background

    Four-dimensional flow cardiovascular magnetic resonance imaging (4D flow CMR) allows comprehensive assessment of pulmonary artery (PA) flow dynamics. Few studies have characterized longitudinal changes in pulmonary flow dynamics and right ventricular (RV) recovery following a pulmonary endarterectomy (PEA) for patients with chronic thromboembolic pulmonary hypertension (CTEPH). This can provide novel insights of RV and PA dynamics during recovery. We investigated the longitudinal trajectory of 4D flow metrics following a PEA including velocity, vorticity, helicity, and PA vessel wall stiffness.


    Twenty patients with CTEPH underwent pre-PEA and > 6 months post-PEA CMR imaging including 4D flow CMR; right heart catheter measurements were performed in 18 of these patients. We developed a semi-automated pipeline to extract integrated 4D flow-derived main, left, and right PA (MPA, LPA, RPA) volumes, velocity flow profiles, and secondary flow profiles. We focused on secondary flow metrics of vorticity, volume fraction of positive helicity (clockwise rotation), and the helical flow index (HFI) that measures helicity intensity.


    Mean PA pressures (mPAP), total pulmonary resistance (TPR), and normalized RV end-systolic volume (RVESV) decreased significantly post-PEA (P < 0.002). 4D flow-derived PA volumes decreased (P < 0.001) and stiffness, velocity, and vorticity increased (P < 0.01) post-PEA. Longitudinal improvements from pre- to post-PEA in mPAP were associated with longitudinal decreases in MPA area (r = 0.68, P = 0.002). Longitudinal improvements in TPR were associated with longitudinal increases in the maximum RPA HFI (r=-0.85, P < 0.001). Longitudinal improvements in RVESV were associated with longitudinal decreases in MPA fraction of positive helicity (r = 0.75, P = 0.003) and minimum MPA HFI (r=-0.72, P = 0.005).


    We developed a semi-automated pipeline for analyzing 4D flow metrics of vessel stiffness and flow profiles. PEA was associated with changes in 4D flow metrics of PA flow profiles and vessel stiffness. Longitudinal analysis revealed that PA helicity was associated with pulmonary remodeling and RV reverse remodeling following a PEA.

    more » « less