SRAM-Based CIM Architecture Design for Event Detection.

Convolutional neural networks (CNNs) play a key role in deep learning applications. However, the high computational complexity and high-energy consumption of CNNs trammel their application in hardware accelerators. Computing-in-memory (CIM) is the technique of running calculations entirely in memory (in our design, we use SRAM). CIM architecture has demonstrated great potential to effectively compute large-scale matrix-vector multiplication. CIM-based architecture for event detection is designed to trigger the next stage of precision inference. To implement an SRAM-based CIM accelerator, a software and hardware co-design approach must consider the CIM macro's hardware limitations to map the weight onto the AI edge devices. In this paper, we designed a hierarchical AI architecture to optimize the end-to-end system power in the AIoT application. In the experiment, the CIM-aware algorithm with 4-bit activation and 8-bit weight is examined on hand gesture and CIFAR-10 datasets, and determined to have 99.70% and 70.58% accuracy, respectively. A profiling tool to analyze the proposed design is also developed to measure how efficient our architecture design is. The proposed design system utilizes the operating frequency of 100 MHz, hand gesture and CIFAR-10 as the datasets, and nine CNNs and one FC layer as its network, resulting in a frame rate of 662 FPS, 37.6% processing unit utilization, and a power consumption of 0.853 mW.

Assessment of left atrial function by full volume real-time three-dimensional echocardiography and left atrial tracking in essential hypertension patients with different patterns of left ventricular geometric models.

OBJECTIVE: To evaluate left atrial function in essential hypertension patients with different patterns of left ventricular geometric models by real time three-dimensional echocardiography (RT-3DE) and left atrial tracking (LAT). METHODS: One hundred and thirty-three patients with essential hypertension were divided into four groups according to criteria proposed by Ganau et al: 36 patients in normal pattern group (I), 34 patients in concentric remodelling pattern group (II), 33 patients in concentric hypertrophy pattern group (III), and 30 patients in eccentric hypertrophy pattern group (IV). Thirty-two healthy volunteers were enrolled in this study. RT-3DE and LAT were used to obtain the volume-time and emptying fraction curves, and then the maximal volume, minimal volume, stroke volume, presystolic volume, ejection fraction, systolic filling rate, early diastolic emptying rate and late diastolic emptying rate of the left atrium were measured. Conduit volume and passive emptying volume of the left atrium were calculated. RESULTS: Left atrial indexes, conduit volume, passive emptying volume, and early diastolic emptying rate decreased, while the other indices increased in the group I compared with the control group (all P<0.05). Compared with the control group, I and II groups, the left atrial volume parameters manifested change in the group III and group IV (all P<0.01). Left atrial ejection fraction measured by RT-3DE or LAT, and late diastolic emptying rate were lower in the group IV compared with the group III (all P<0.05). There was no difference in left atrial conduit volume, passive emptying volume, and early diastolic left atrial emptying rate between the group II and group I (all P>0.05). There was a positive correlation between RT-3DE and LAT in evaluating left atrial function. CONCLUSIONS: Left atrial booster pump function decreases in the eccentric hypertrophy pattern. Left atrial conduit function is unchanged in the normal pattern and the concentric remodelling pattern groups. RT-3DE and LAT technology can be used for quantitative evaluation of left atrial volume and function.