Method details

Download PDF file of article
Probabilistic Model Based Evaluation of Coronary Artery Stenosis on Computed Tomography Angiography



Abstract:
For evaluating the severity of luminal stenosis of coronary artery, we propose a fully automated probabilistic approach in detecting and quantifying coronary artery plaque which serves as evidence of luminal narrowing of as-sessed degree in three processing steps. First, a Gaussian distribution is utilized to adaptively allocate the center of cross-sectional plaque with the variance of posterior density as the plausible size. A concentric model fixed to the vessel centerline is applied to roughly segment the vessel lumen in the same fashion as plaque detection. Second, the quantitative evaluation of diameter stenosis is de-termined using the Kalman filtering formulation. Third, the stenosis degree is given using Bayes classifier based on the posterior probability of severity con-ditioned on stenosis percentage and plaque type of the training data. The meth-od was evaluated on the 42 challenging dataset of Rotterdam Coronary Artery Algorithm Evaluation Framework. For the detection, a sensitivity of 0.54 and a PPV of 0.13 is obtained as compared to QCA, while a sensitivity of 0.29 and a PPV of 0.03 is achieved as compared to CTA. The stenosis degree is estimated with an absolute average difference of 38.5% as compared to QCA, and a weighted kappa value of -0.04 is obtained as compared to CTA.

Detection confusion tables

Calc. cat.QCA (per segment)CTA (per lesion)
 TPFPFNTP+FPTP+FNTPFPFNTP+FPTP+FN
All  14  87  14  101  28  15  484  32  499  47 
The results of this method are based on the following centerlines: LKEB team manual.

Detection (QCA per segment / CTA per lesion)

Calc. cat.QCA
Sens.
QCA
P.P.V.
CTA
Sens.
CTA
P.P.V.
Avg. rank
 %rank%rank%rank%rank 
0 (0 - 0.1)  20.0  11.0  7.1  13.0  14.3  14.0  0.9  17.0  13.8 
1 (0.1 - 10)  100.0  1.0  7.7  11.0  40.0  11.0  3.2  15.0  9.5 
2 (11 - 100)  71.4  4.0  18.5  11.0  54.5  8.0  4.1  18.0  10.2 
3 (101 - 400)  40.0  13.0  12.5  16.0  21.1  14.0  3.1  19.0  15.5 
4 (400+)  60.0  4.0  20.0  14.0  40.0  11.0  4.1  19.0  12.0 
All  50.0  12.0  13.9  14.0  31.9  14.0  3.0  19.0  14.8 
For ranking  50.0  12.0  13.9  14.0  31.9  14.0  3.0  19.0  14.8 
These results are based on 30 datasets and 19 submissions.


Quantification

Calc. cat.QCA
Avg. Abs. diff.
QCA
R.M.S. diff.
CTA
Weigthed Kappa
Avg. rank
 %rank%rankΚrank 
0 (0 - 0.1)  35.4  9.0  38.6  7.0  -0.01  15.0  11.5 
1 (0.1 - 10)  37.3  12.0  40.8  12.0  -1.00  15.0  13.5 
2 (11 - 100)  38.3  9.0  42.2  9.0  -1.00  15.0  12.0 
3 (101 - 400)  40.7  9.0  45.8  9.0  -0.05  15.0  12.0 
4 (400+)  40.9  10.0  44.6  9.0  -0.02  15.0  12.2 
All  38.6  9.0  42.7  9.0  -0.03  15.0  12.0 
For ranking  38.6  9.0  42.7  9.0  -0.03  15.0  12.0