P-007 Strategic AI integration in endovascular therapy: a comparative analysis of ChatGPT, google bard, and AtlasGPT in intracranial aneurysm decision-making

Abstract

<h3>Objective</h3> This study aims to elucidate the impact of Artificial Intelligence (AI) on the field of neurointervention and endovascular therapy, offering insights into the future of AI-assisted medical practices and their potential to revolutionize patient care in complex, high-stakes medical domains. <h3>Methods</h3> Twenty different concise clinical cases of endovascular treatment of intracranial aneurysms were presented to four different AIs (ChatGPT 3.5, ChatGPT 4.0. Google Bard and AtlasGPT), two neurosurgery residents and one interventional neuroradiology (INR) clinical fellow. Their responses were evaluated by four endovascular treatment specialists, who judged blindly the correctness and the quality of decision of the treatment options. <h3>Results</h3> In both assessments, AtlasGPT presented the best rates of all AIs (85–90% correctness and 8–9 in quality of decision), being comparable to the INR fellow and even surpassing the performance of the neurosurgery residents. ChatGPT 4.0 also performed well in the quality of decision assessment. Humans presented a wide arrow of performances, varying from 45–55% in institution A to 85–90% in the INR fellow responses. <h3>Conclusion</h3> This study highlights the great current decision making potential of AI in complex cases of endovascular treatments for intracranial aneurysms, reinforcing its role as a powerful tool to help medical professionals in their daily activities, despite its clear limitations. <h3>Disclosures</h3> <b>S. Batista:</b> None. <b>E. Tanus:</b> None. <b>R. Camerotte:</b> None. <b>C. Alves Filho:</b> None. <b>F. Braga:</b> None. <b>P. Lopes:</b> None. <b>G. Galvão:</b> None. <b>G. Vellasques:</b> None. <b>J. Almeida Filho:</b> None. <b>P. Pereira:</b> None. <b>P. Niemeyer Filho:</b> None.

Ähnliche Arbeiten