Eur Radiol: The risk of rupture of intracranial aneurysms is predicted by radiology! 

Intracranial aneurysms are present in 3-7% of the population, and although the rate of rupture of aneurysms is low, the mortality rate of aneurysm rupture is as high as 65%, so it is important to identify an aneurysms with a tendency to rupture.
morphology and hemodynamics of these aneurysms have been shown to be associated with rupture.
, such as high blood pressure, blood lipid levels, alcohol consumption, smoking, etc. are also risk factors for rupture of aneurysms.
based on these risk factors, various methods of assessing risk have been proposed clinically, but predictive performance still needs to be improved.
stage, artificial intelligence has been applied to the detection, risk management and treatment of intracranial aneurysms but is being developed.
radiology is a new analytical technique, which extracts features from images and analyzes them by data-driven methods.
it has been shown to be effective in many fields such as cancer prognostication, radiation therapy, tumor genetics assessment, etc.
, however, the application of radiology in predicting aneurysm rupture has yet to be explored.
recently, a study published in the journal European Radiology applied radiology to assess the difference between ruptured and unbromed aneurysms and explored their potential application value in predicting aneurysm rupture.
study included 122 aneurysms (93 unbromed aneurysms).
extraction of morphological and radiologic characteristics for each case.
statistical analysis to determine significant characteristics and incorporate them into predictive models built by machine learning algorithms.
in order to study the effectiveness of radiologic characteristics, we constructed and compared three models.
baseline model A is constructed with morphological features, model B adds radiological shape features, and model C adds more radiological features.
analysis of the 10 most important variables in model C to determine independent risk factors.
established a simplified model based on independent risk factors for clinical use.
5 morphological features and 89 radiological features were significantly associated with aneurysm rupture.
of the recipient working characteristic curves of models A, B and C are 0.767, 0.807 and 0.879, respectively.
model C is significantly better than models A and B (p slt; 0.001).
analysis found that the two radiological features used to build simplified models showed AUROC of 0.876.
the ROC curve (a) and the precision recall curve (b) for models a, model b, and model c.
model B consists of morphological features and radiological group morphological features.
model C consists of morphological features, radiation group shape features, first-order histogram features, and second-order texture features.
point on the curve indicates that the truncated values determined by the Jordon index are different in the radiological characteristics of ruptured aneurysms and unbromed aneurysms.
use of radiological features, especially texture characteristics, can significantly improve the performance of predicting aneurysm rupture.