Two-year experience of stroke thrombectomy referral pathway and use of AI imaging support

<h3>Introduction</h3> Thrombectomy involves mechanical removal of large vessel occlusion (LVO) from proximal circulation in acute ischaemic stroke. Rapid diagnosis of LVO with multiphase CT angiography (mCTA) using AI can facilitate timely referral. <h3>Methods</h3> We examined patients considered for thrombectomy who underwent mCTA with Brainomix’s e-CTA analysis and/or were referred for thrombectomy. We collected clinical and imaging characteristics, door-to-needle (DTN), door-in-door-out (DIDO), and stroke onset-to-DSA times. <h3>Results</h3> We identified 59 referred and 59 non-referred patients. The referred patients had longer median time from CT to CTA (p&lt;0.05). Among referred patients, 37% were transferred for thrombectomy, 24% (fourteen) received thrombectomy, and 14% had recanalised by arrival. Median NIHSS change 24 hours after thrombectomy was -3 (-13 to - 2). Median DIDO time was 104 (87-126) minutes, and stroke onset to DSA time was 308 (261-505) minutes. e-CTA analysis exhibited sensitivity of 0.78 (95% CI: 0.56-0.93) and specificity of 0.96 (0.87-1.00) for identifying LVO. Patients receiving mCTA had a shorter median DTN time than patients who were thrombolysed but did not receive mCTA. <h3>Conclusion</h3> 12% of patients assessed for referral received thrombectomy. mCTA does not delay thrombolysis. Factors that may improve thrombectomy provision include reducing delay to mCTA, DIDO, and expanding referral time window.