Accuracy and reliability of computer-assisted semi-automated morphological analysis of intracranial aneurysms: an experimental study with digital phantoms and clinical aneurysm cases.

Abstract

Morphological parameters are very important for predicting aneurysm rupture. However, due to geometric radiographic distortion and plane/angle selection bias, the traditional manual measurements (MM) of aneurysm morphology are inaccurate and suffer from severe variability. Our study is to evaluate the accuracy and reliability of computer-assisted semi-automated measurement (CASAM) of intracranial aneurysms, which is a novel technique in aneurysm measurement. An in-house software for CASAM was developed. Classical morphology indices including aneurysm diameter, neck size, height, width, volume, inflow angle, and aspect ratio were measured. To validate the accuracy and robustness of the semi-automated measurements, 20 digital intracranial aneurysm phantoms and 27 clinical aneurysms with different locations and sizes were measured using MM or CASAM. In the phantom study, although the inter-observer variability of both the MM and CASAM was very low, the manual measurements had higher errors (1.7-19.1%), while the CASAM yielded more accurate results (errors of 1.1-2.5%). The consistency test indicated that the CASAM results were highly consistent with the actual values (concordance correlation coefficient = 0.993). In the clinical study, CASAM showed better intraclass correlation coefficient values compared with MM. The inflow angle had low consistency in both groups. We successfully developed a computer-assisted method to semi-automatically measure the morphological parameters of aneurysm. According to our study, CASAM of aneurysm morphological parameters is a more precise and reliable way than MM to obtain accurate aneurysm morphological parameters. This method is worthy of further studies to promote its clinical use.