Search

Background: Continuous‐flow ventricular assist devices (cfVADs) are implanted in patients with end‐stage heart failure to assist with blood circulation. However, VAD implantation is associated with dangerous thrombotic complications. Our goal was to determine the impact of micron and sub‐micron scale Ti6Al4V surface roughness on adherent platelet aggregate properties under clinically relevant shear rates. Methods: We used fluorescence microscopy to visualize platelets in real time as they adhered to Ti6Al4V coupons of varying degrees of roughness, including a smooth control, in microfluidic channels and quantified deposition using an image processing algorithm. We systematically characterized roughness using spatial frequencies to generalize results for more blood‐biomaterial contact applications. Results: We observed that on the control and sub‐micron rough surfaces, at 1000 s−1, platelets adhered uniformly on the surface. At 2000 s−1, we observed small and stably adherent platelet aggregates. At 5500 s−1, platelet aggregates were large, unstable and interconnected via fibrillar structures. On a surface with micron‐scale roughness features, at all three shear rates, platelets deposited in the troughs of the roughened surface, and formed aggregates. Thrombus height at 2000 s−1 and 5500 s−1 was greatest on the roughest surface and lowest on the mirror‐finished surface, as indicated by the mean fluorescence intensity. Conclusions: These results demonstrated that at high shear rates, thrombi form regardless of surface topography at the scales applied. At lower shear rates, micron‐scale surface features cause thrombus formation, whereas submicron features result in innocuous platelet adhesion. These findings have implications for manufacturing costs and other considerations. [ABSTRACT FROM AUTHOR]

Studies on deep-learning-based behavioral pattern recognition have recently received considerable attention. However, if there are insufficient data and the activity to be identified is changed, a robust deep learning model cannot be created. This work contributes a generalized deep learning model that is robust to noise not dependent on input signals by extracting features through a deep learning model for each heterogeneous input signal that can maintain performance while minimizing preprocessing of the input signal. We propose a hybrid deep learning model that takes heterogeneous sensor data, an acceleration sensor, and an image as inputs. For accelerometer data, we use a convolutional neural network (CNN) and convolutional block attention module models (CBAM), and apply bidirectional long short-term memory and a residual neural network. The overall accuracy was 94.8% with a skeleton image and accelerometer data, and 93.1% with a skeleton image, coordinates, and accelerometer data after evaluating nine behaviors using the Berkeley Multimodal Human Action Database (MHAD). Furthermore, the accuracy of the investigation was revealed to be 93.4% with inverted images and 93.2% with white noise added to the accelerometer data. Testing with data that included inversion and noise data indicated that the suggested model was robust, with a performance deterioration of approximately 1%. [ABSTRACT FROM AUTHOR]