The healthcare sector is undergoing a transformative phase, driven by the integration of Artificial Intelligence (AI) into diagnostic and patient care processes. A significant area of interest within this technological evolution is the application of AI in predicting the rupture risk of brain aneurysms, a condition affecting approximately 3.2% of the global population. Brain aneurysms, characterised by bulges in cerebral blood vessels, present significant management challenges due to the severe consequences of rupture and the inherent risks of surgical intervention.
A recent comprehensive review has investigated the efficacy of AI models in assessing the risk of aneurysm rupture, analysing 20 studies and over 20,000 individual cases. The results demonstrated that AI models achieved accuracy rates ranging from 66% to 90%, which are comparable to or slightly superior to traditional methods such as the PHASES score. However, the variability among the studies rendered it difficult to draw definitive conclusions, prompting the authors to caution that AI models are not yet ready for clinical implementation.
Current applications of AI in aneurysm rupture prediction are particularly promising, especially with the utilisation of machine learning (ML) and deep learning (DL) techniques. Traditional ML algorithms, such as random forests and support vector machines, are structured around predefined frameworks informed by expert knowledge. Conversely, DL models employ artificial neural networks to autonomously extract data from images, creating feature maps and learning visual patterns. These advancements suggest a potential paradigm shift in risk assessment, though the path to clinical application is fraught with challenges.
The management of intracranial aneurysms (IAs) remains complex, requiring physicians to balance the risks of intervention against the potential for subarachnoid haemorrhage. The heterogeneity in clinical practice is evident in the management of small unruptured aneurysms (UIAs), with considerable differences in the frequency and modality of follow-up imaging. A survey among neurointerventionalists and neuroradiologists indicated that 84% recommended annual imaging surveillance for UIAs, with a preference for noncontrast MR angiography. Early detection and precise prediction of aneurysm rupture risk are crucial to avoiding catastrophic outcomes and unnecessary interventions.
Despite the promise of AI models in addressing these challenges through more accurate and evidence-based risk assessments, limitations persist. Current models require further validation and enhancement in precision. Several studies have highlighted AI’s potential in predicting aneurysm rupture. For example, a study employing a two-layer feedforward artificial neural network (ANN) achieved a prediction accuracy of 94.8% for 594 aneurysms. However, the study’s limitations, including a small sample size and lack of long-term follow-up data, underscore the necessity for further research and validation.
Another investigation, using data from Korea’s National Health Screening Program, developed predictive models with accuracy rates between 75% and 77%. While these results are promising, the models necessitate further refinement and testing across diverse populations to ensure their effectiveness as screening tools. Additionally, the potential of convolutional neural networks (CNNs) in predicting aneurysm rupture is noteworthy. A study employing a multi-view CNN based on 3D-DSA achieved an overall accuracy of 81.72%, surpassing human assessors in diagnostic precision. Nonetheless, the authors emphasise the need for more extensive, data-driven research to enhance diagnostic accuracy and facilitate practical application.
This systematic review highlights the significant potential of AI in revolutionising the prediction of brain aneurysm rupture risk. Although current models show considerable promise, ongoing research and validation are crucial to ensuring their clinical applicability and reliability. As AI continues to evolve, it holds the potential to transform the management of intracranial aneurysms, thereby improving patient outcomes and optimising healthcare resources. The journey towards integrating AI into routine clinical practice appears promising, yet challenging, necessitating a concerted effort from researchers, clinicians, and technologists alike.
Be the first to comment