Effect of Geometric-Based Coronary Calcium Volume as a Feature Along with its Shape-Based Attributes for Cardiological Risk Prediction from Low Contrast Intravascular Ultrasound

The assessment and study of sensitivity analysis of cardiological risk using coronary calcium quantification in patients with increased neurovascular risk is considerably enhanced by taking normalized calcium volume and shape-based coronary lesion characteristics as features. Intravascular ultrasound images were acquired from 92 patients with stable angina pectoris. The risk quantification of each patient was performed by computing different shape-based features of the coronary calcium lesions, namely: (i) mean lesion thickness, (ii) mean standard deviation of lesion thickness, (iii) mean centreline lesion length, (iv) mean length of all lesion branches, (v) mean span (arc angle) of the detected lesion, (vi) mean lesion irregularity, and (vii) lesion distance to catheter center. For normalized calcium volume in IVUS coronary arteries, we adapted a geometric-based segmentation strategy by suppressing the non-calcium region thereby isolating the calcium lesion. Our results demonstrate that the area under the curve (AUC) was 0.616 when combining volume feature with shape-based features in comparison to 0.58 when using shape-based features alone. Receiver operating characteristic (ROC) curve analysis shows that the statistical significance of the established association helps in analysing the sensitivity of coronary calcium quantification in neurological risk patients. The analysis reports show an improvement in AUC by 6.2% when using the combination of volume and shape-based features compared to shape-based features alone