scholarly journals Continuous, Full-scope, Spatio-temporal Tracking Metric based on KL-divergence

Author(s):  
Terry Adams
Circulation ◽  
2015 ◽  
Vol 131 (suppl_2) ◽  
Author(s):  
Mitchel Benovoy ◽  
Farida Cheriet ◽  
Roch L Maurice ◽  
Nagib Dahdah

Background: Mechanical properties of coronary arteries (CA) hold clues to vascular health and viability. Traditionally assessed with intracoronary imaging, we present an angiography-based system to assess CA vasomotion using automatic vessel segmentation and spatio-temporal tracking. Elastic moduli computed from dynamic CA calibers are compared between non-KD patients (CTL), KD patients with no CA aneurysms (KDAN-), and those with aneurysms (KDAN+). Methods: Proximal CA angiograms are automatically segmented and tracked over a cardiac cycle. CA centerline is extracted and the mean caliber is computed from diameters along its length. The resulting caliber variation reflects the CA vasomotion (Figure 1a). We then calculated the Vasomotion Standard Deviation (VSD) and CA recoil with the mean constriction velocity (MCV). Finally, Elastic Pressure moduli were computed using trans-myocardium pressure gradients. Results: We analyzed 51 left CA segments from 23 patients (5 CTL, 5 KDAN-, 13 KDAN+). Data are mean ± SD normalized pixels (npx). VSD was significantly reduced ( p <0.01) in KDAN+ (0.25±0.05) and KDAN- (0.27±0.04) vs CTL (0.38±0.07 npx). Coronary recoil was significantly reduced (p<0.05) in KDAN+ vs CTL, with MCV 3.50±0.67 vs 4.59±1.94 npx/sec. Pressure-dependent stiffness characteristics were equally atypical (Figure 1b). Conclusion: The proposed angiography-based stiffness assessment system shows abnormal CA vascular physiology in our cohort of KD patients. These results concur with previous invasive studies. The potential usability of this system for vascular health assessment could be applied to previously recorded CA angiograms for risk stratification.


2018 ◽  
Vol 463-464 ◽  
pp. 56-74 ◽  
Author(s):  
Adrian N. Bishop ◽  
Jeremie Houssineau ◽  
Daniel Angley ◽  
Branko Ristic

1999 ◽  
Vol 18 (10) ◽  
pp. 957-972 ◽  
Author(s):  
J. Huang ◽  
D. Abendschein ◽  
V.G. Davila-Roman ◽  
A.A. Amini

2009 ◽  
Vol 3 (5) ◽  
pp. e448 ◽  
Author(s):  
Adriano Mondini ◽  
Roberta Vieira de Moraes Bronzoni ◽  
Silvia Helena Pereira Nunes ◽  
Francisco Chiaravalloti Neto ◽  
Eduardo Massad ◽  
...  

2020 ◽  
Vol 6 (5) ◽  
pp. 27
Author(s):  
Andrew Tzer-Yeu Chen ◽  
Morteza Biglari-Abhari ◽  
Kevin I-Kai Wang

Knowing who is where is a common task for many computer vision applications. Most of the literature focuses on one of two approaches: determining who a detected person is (appearance-based re-identification) and collating positions into a list, or determining the motion of a person (spatio-temporal-based tracking) and assigning identity labels based on tracks formed. This paper presents a model fusion approach, aiming towards combining both sources of information together in order to increase the accuracy of determining identity classes for detected people using re-ranking. First, a Sequential k-Means re-identification approach is presented, followed by a Kalman filter-based spatio-temporal tracking approach. A linear weighting approach is used to fuse the outputs from these models together, with modification of the weights using a decay function and a rule-based system to reflect the strengths and weaknesses of the models under different conditions. Preliminary experimental results with two different person detection algorithms on an indoor person tracking dataset show that fusing the appearance and spatio-temporal models significantly increases the overall accuracy of the classification operation.


Sign in / Sign up

Export Citation Format

Share Document