398 Towards Catheter-Clot Engagement Detection Via Classification of Vacuum and Motion Signature During Aspiration Thrombectomy

Abstract

INTRODUCTION: While acute ischemic stroke caused by large vessel occlusion is effectively treated with aspiration thrombectomy, procedural success depends on the quality of engagement between the catheter tip and the thrombus. Currently, there is no method to identify whether the catheter tip is truly in contact with the clot surface. Poor thrombus engagement can result in failure of recanalization. METHODS: An in vitro arterial occlusion was modeled with Tygon tubing representing the blood vessel and a synthetic thrombus made from TrueClot simulants. A 6Fr microcatheter was controllably inserted into the mock vessel until it contacted the synthetic thrombus. A motorized syringe and pressure sensor were connected proximally to the catheter to precisely control and measure the cyclic vacuum excitation in the catheter. The cyclic vacuum excitation was performed with a stroke of ±0.5 mm at 2 Hz, 4 Hz, and 8 Hz at catheter-clot distances of 150 mm, 20 mm, 10 mm, 5 mm, and in contact on 17 different clot models. Control and sensing data was collected at 1000 Hz frequency. For each catheter-clot distance, 1000 vacuum measurements were collected and pressure traces were saved. The change in the pressure mean per a change in insertion depth of the catheter was used to train a support vector machine (SVM) classifier that confirms clot contact. RESULTS: Using 30% of the clot model samples to train the SVM classifier and the remaining samples for validation, the quality of classification was determined. The accuracy (and F1 score) of clot detection at 2 Hz, 4 Hz, and 8 Hz were 98.31% (95.65%), 100% (100%), and 93.22% (80.00%) respectively. CONCLUSIONS: Accurate detection of catheter-clot contact is feasible via small excitation of the aspiration syringe and concurrent measurement of vacuum pressure.