Method details

Download PDF file of article
Coronary Artery Stenoses Detection with Random Forest



Abstract:
The paper describes a method solving stenosis detection problem for the coronary artery stenosis detection and quanti cation challenge. The adopted solution uses rotation invariant features extracted around predefi ned points along artery centreline. The stenoses detection is obtained using random forest classifi er. The paper reports on the results obtained on the training and test data sets with di erent values of the random forest parameters. On the test data, for the detection, a sensitivity of 61% and PPV of 15% is obtained as compared to QCA, while a sensitivity of 29% and PPV of 8% is achieved as compared to CTA. The future work to improve the detection results is also briefly introduced.

Detection confusion tables

Calc. cat.QCA (per segment)CTA (per lesion)
 TPFPFNTP+FPTP+FNTPFPFNTP+FPTP+FN
All  16  115  12  131  28  20  243  27  263  47 
The results of this method are based on the following centerlines: Rcadia team auto.

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)  60.0  4.0  13.0  10.0  28.6  11.0  3.5  15.0  10.0 
1 (0.1 - 10)  100.0  1.0  8.3  10.0  60.0  6.0  14.3  13.0  7.5 
2 (11 - 100)  57.1  8.0  11.4  14.0  54.5  8.0  7.9  14.0  11.0 
3 (101 - 400)  60.0  8.0  15.0  15.0  36.8  12.0  9.5  16.0  12.8 
4 (400+)  40.0  12.0  9.5  18.0  40.0  11.0  5.7  18.0  14.8 
All  57.1  7.0  12.2  15.0  42.6  12.0  7.6  16.0  12.5 
For ranking  57.1  7.0  12.2  15.0  42.6  12.0  7.6  16.0  12.5 
These results are based on 30 datasets and 19 submissions.