Continuous blood pressure (BP) monitoring using wearable devices has received increasing attention due to its importance in diagnosing diseases. However, existing methods mainly measure BP intermittently, involve some form of user effort, and suffer from insufficient accuracy due to sensor properties. In order to overcome these limitations, we study the BP measurement technology based on heart sounds, and find that the time interval between the first and second heart sounds (TIFS) of bone-conducted heart sounds collected in the binaural canal is closely related to BP. Motivated by this, we propose HearBP, a novel BP monitoring system that utilizes inear microphones to collect bone-conducted heart sounds in the binaural canal. We first design a noise removing method based on U-net autoencoder-decoder to separate clean heart sounds from background noises. Then, we design a feature extraction method based on shannon energy and energy-entropy ratio to further mine the time domain and frequency domain features of heart sounds. In addition, combined with the principal component analysis algorithm, we achieve feature dimension reduction to extract the main features related to BP. Finally, we propose a network model based on dendritic neural regression to construct a mapping between the extracted features and BP. Extensive experiments with 41 participants show the average estimation error of 0.97mmHg and 1.61mmHg and the standard deviation error of 3.13mmHg and 3.56mmHg for diastolic pressure and systolic pressure, respectively. These errors are within the acceptable range specified by the FDA’s AAMI protocol.

Vision-Language Models (VLMs) are expected to be capable of reasoning with commonsense knowledge as human beings. One example is that humans can reason where and when an image is taken based on their knowledge. This makes us wonder if, based on visual cues, Vision-Language Models that are pre-trained with large-scale image-text resources can achieve and even surpass human capability in reasoning times and location. To address this question, we propose a two-stage Recognition & Reasoning probing task applied to discriminative and generative VLMs to uncover whether VLMs can recognize times and location-relevant features and further reason about it. To facilitate the studies, we introduce WikiTiLo, a well-curated image dataset compromising images with rich socio-cultural cues. In extensive evaluation experiments, we find that although VLMs can effectively retain times and location-relevant features in visual encoders, they still fail to make perfect reasoning with context-conditioned visual features. The dataset is available at https://github.com/gengyuanmax/WikiTiLo.

Temporal graphs offer more accurate modeling of many real-world scenarios than static graphs. However, neighbor aggregation, a critical building block of graph networks, for temporal graphs, is currently straightforwardly extended from that of static graphs. It can be computationally expensive when involving all historical neighbors during such aggregation. In practice, typically only a subset of the most recent neighbors are involved. However, such subsampling leads to incomplete and biased neighbor information. To address this limitation, we propose a novel framework for temporal neighbor aggregation that uses the recurrent neural network with node-wise hidden states to integrate information from all historical neighbors for each node to acquire the complete neighbor information. We demonstrate the superior theoretical expressiveness of the proposed framework as well as its state-of-the-art performance in real-world applications. Notably, it achieves a significant +9.4% improvement on averaged precision in a real-world Ecommerce dataset over existing methods on 2-layer models.

Diabetic retinopathy (DR) is a common retinal disease that leads to blindness. For diagnosis purposes, DR image grading aims to provide automatic DR grade classification, which is not addressed in conventional research methods of binary DR image classification. Small objects in the eye images, like lesions and microaneurysms, are essential to DR grading in medical imaging, but they could easily be influenced by other objects. To address these challenges, we propose a new deep learning architecture, called BiRA-Net, which combines the attention model for feature extraction and bilinear model for fine-grained classification. Furthermore, in considering the distance between different grades of different DR categories, we propose a new loss function, called grading loss, which leads to improved training convergence of the proposed approach. Experimental results are provided to demonstrate the superior performance of the proposed approach.