Non-invasive Detection Of Reduced Leaflet Motion In Transcatheter Aortic Valves Via Embedded Wireless Pressure Microsensors - An In-silico Proof-of-concept
Shantanu Bailoor1, Jung Hee Seo1, Lakshmi P. Dasi2, Stefano Schena3, Rajat Mittal1
1Johns Hopkins University, Baltimore, MD, USA, 2Georgia Institute of Technology, Atlanta, GA, USA, 3Johns Hopkins Medical Institute, Baltimore, MD, USA.
OBJECTIVE: Leaflet thrombosis can cause unexpected, premature failure of transcatheter aortic valves (TAVs). Patients may benefit from longitudinal screening for reduced leaflet motion (RLM). We present in-silico proof-of-concept and evaluation of a novel monitoring modality that employs embedded wireless pressure microsensors on TAV stents for early, non-invasive detection of RLM.
METHODS: 3D ascending aorta models were segmented from CT-scans of 6 patients screened to receive AV replacement. TAVs were installed in-silico in these models and computational fluid dynamics simulations were performed with varying levels of RLM on TAV leaflets (N=126 leaflets). Hemodynamic pressure was recorded at three axial locations per leaflet on the stent: skirt, sinus, and sino-tubular junction. Trans-leaflet (ΔpL=pskirt-pSTJ) and sinus (ΔpS=pskirt-psinus) gradients were used to define a set of 14 statistical features for each leaflet and ranked using ANOVA F-score. Feature-pairs were created from the same gradient to maximize separability between "Healthy" and "RLM" leaflets.
RESULTS: Peak systolic ΔpL asymmetry (F=243, p<0.001), calculated as the deviation of individual leaflet ΔpL from mean ΔpL over the valve at the instant of maximum ΔpL across leaflet, and the corresponding gradient ΔpL (F=168, p<0.001) was the feature-pair most strongly correlated with individual leaflet status (Figure). A linear discriminant classifier was trained to retrospectively distinguish healthy leaflets from those with RLM in this feature space with 99.2% accuracy (1 false positive). Prospective testing was performed on 7 simulations from a new patient anatomy, with 12 Healthy, 9 RLM leaflets (N=21). All tested leaflets were accurately classified.
CONCLUSIONS: Peak ΔpL recorded on TAV stents and its corresponding asymmetry, coupled with machine learning can detect RLM in individual leaflets. This provides proof-of-concept that advanced TAV designs with embedded wireless pressure sensors could enable non-invasive early RLM detection.
Back to 2022 Abstracts