Bozsak, Franz

Background Large vessel occlusion (LVO) stroke is associated with a high risk of permanent disability. There are no reliable tools providing periprocedural information on clot characteristics, such as composition and length, that may impact first pass effect, which has been shown to improve functional outcomes. In the first study of a novel technology, we evaluated whether the Clotild Smart Guidewire System (Clotild), an impedance-based device, can provide in situ information on clot characteristics during a mechanical thrombectomy procedure. Methods This is the first-in-human, international, multicentric, single-arm trial involving LVO stroke patients. Of 45 enrolled subjects, 41 constituted the safety population; 39 was the per protocol population, of which 26 subjects with complete impedance data comprised the performance population. The latter was divided into development (11 subjects) and validation (15 subjects) populations for prediction model development and assessment. The safety endpoint assessed Clotild-related intracranial vessel perforation/dissection. Performance endpoints evaluated Clotild’s ability to perform classification of impedance measurements distinguishing substantial versus negligible local content in red blood cells (RBCs) and platelets, to detect the arterial wall and the distal end of the clot. Results The safety endpoint was met in all subjects. Performance analysis comparing the agreement between Clotild’s output and impedance experts’ readings showed an area under the receiver operating characteristic curve (AUC) of 0.97 (95% CI 0.93 to 0.99) for RBC, and 0.94 (95% CI 0.89 to 0.98) for platelet clot composition. Regarding detection, Clotild’s performance showed an AUC of 0.69 (95% CI 0.59 to 0.77) for the arterial wall, and 0.83 (95% CI 0.59 to 1.00) for the distal end of the clot. Conclusions Clotild was proven safe to acquire impedance data and to provide information on in situ clot characteristics in large vessel occlusions.

Background Intra-procedural characterization of stroke thromboemboli might guide mechanical thrombectomy (MT) device choice to improve recanalization rates. Electrochemical impedance spectroscopy (EIS) has been used to characterize various biological tissues in real time but has not been used in thrombus. Objective To perform a feasibility study of EIS analysis of thrombi retrieved by MT to evaluate: (1) the ability of EIS and machine learning to predict red blood cell (RBC) percentage content of thrombi and (2) to classify the thrombi as “RBC-rich” or “RBC-poor” based on a range of cutoff values of RBC. Methods ClotbasePilot was a multicentric, international, prospective feasibility study. Retrieved thrombi underwent histological analysis to identify proportions of RBC and other components. EIS results were analyzed with machine learning. Linear regression was used to evaluate the correlation between the histology and EIS. Sensitivity and specificity of the model to classify the thrombus as RBC-rich or RBC-poor were also evaluated. Results Among 514 MT,179 thrombi were included for EIS and histological analysis. The mean composition in RBC of the thrombi was 36% ± 24. Good correlation between the impedance-based prediction and histology was achieved (slope of 0.9, R 2 = 0.53, Pearson coefficient = 0.72). Depending on the chosen cutoff, ranging from 20 to 60% of RBC, the calculated sensitivity for classification of thrombi ranged from 77 to 85% and the specificity from 72 to 88%. Conclusion Combination of EIS and machine learning can reliably predict the RBC composition of retrieved ex vivo AIS thrombi and then classify them into groups according to their RBC composition with good sensitivity and specificity.