Uncertainty measurement and confidence calibration for calcium detection in optical coherence images

Optical coherence tomography (OCT) has become increasingly essential in assisting the treatment of coronary artery disease (CAD). Image-guided solutions such as Percutaneous Coronary Intervention (PCI) are extensively used during the treatment of CAD. However, unidentified calcified regions within a narrowed artery could impair the outcome of the PCI. Prior to treatments, object detection of the diseased regions is paramount to automatically procure accurate readings on calcifications within the artery. Though deep learning-based object detection methods have been explored in a variety of applications, the quality of predictions can be negatively impacted by overconfident deep learning models, which is not desirable in safety-critical scenarios. In this work, we adopt an object detection model to rapidly draw the calcified region from coronary OCT images using bounding box. We evaluate the uncertainty of predictions based on the expected calibration errors, thus assessing the certainty level of detection results. The calibrated confidence of prediction results in a confidence error of approximately 0.13, suggesting that the confidence calibration on calcification detection could provide a more trustworthy result, which indicates a great potential to assist clinical evaluation of treating the CAD during the imaging-guided procedure.