The diagnostic function of the catheterization laboratory is not growing as other modalities have minimized and often replaced the need for invasive hemodynamic and angiographic assessment in all forms of heart disease. The interventional cardiac catheterization lab is growing with coronary, peripheral, carotid, and a myriad of structural heart disease interventions. It is in the context of performing an intervention that traditional invasive diagnostic cardiac catheterization methods survive and warrant enhancements.
The study ''Automated Left Ventricular Delineation in X-ray Angiograms: A Validation Study'' in this issue of CCI presents data on the performance of a different algorithm to automatically determine the borders of the left ventricular cavity during selective catheter injection of contrast. The results show that the new, commercially available algorithm is a bit faster and is more consistent than done by an experienced operator. It should be noted that completely automated analysis is not being studied. There is still manual review of the left ventriculogram, selection of end-diastolic and endsystolic frames, calibration, and apparently choosing several reference points on the contour of the LV before the algorithm performs its border recognition. Editing was then needed on one out of five end-diastolic images and one out of four end-systolic images.
The software was tested on only 30 left ventriculograms, and no details are given as to the variability of LV size, shape, and confounding factors such as mitral regurgitation for an automated contour detection algorithm. Performance of a segmentation algorithm must be judged by the nature of the diversity of the image content. Likewise, all analyses were single plane, and all were in an RAO projection rather than an LAO projection.
Enhanced workflow is a selling point for this software. Compared to manual LV analysis, there is a 26% time reduction using their approach, a decrease from 4.2 to 3.1 min. The impact on clinical practice of such a small reduction will not be as significant as when doing laborious frame-by-frame analysis as done in the past to study