3 May 2024 in Medical Image Analysis
13 February 2025 in IEEE Access
Variability in sonographer expertise often leads to low-quality ultrasound imaging, presenting significant challenges for accurate echocardiogram video segmentation. Current methods require extensive annotations, which are impractical given the large number of frames and artifacts in videos. To address this, we propose a Multi-Cluster Contrastive (MCC) learning framework, a semi-supervised approach that minimizes annotation requirements while maintaining high segmentation performance. Leveraging contrastive loss to enhance foreground feature extraction, our method incorporates multi-cluster contrastive loss to utilize multiple annotated ground-truths per batch and an anchor frame selection algorithm to improve segmentation performance. Experimental results on two public echocardiography datasets (MCE and EchoNet-Dynamic) demonstrate the effectiveness of our method, achieving state-of-the-art performance. The MCC framework enhances segmentation practicality by reducing annotation requirements, particularly for developing new datasets, and facilitates efficient segmentation of low-quality echocardiogram videos. Our implementation is available at https://github.com/windstormer/MCC.
12 June 2026 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Edge devices capable of running artificial intelligence (AI) applications have seen a surge in demand for energy-efficient and high-throughput computation. In this study, a 22-nm edge–AI processor, incorporating an accelerator with error-free Booth-value-confined (BVC) multiprecision (MP) multiplier and near-memory computing (NMC), is introduced to accelerate neural networks (NNs). It has the following three major features. First, a BVC MP multiplier based on radix-8 Booth (R8B) is introduced to reduce computation complexity by prohibiting the “±3” cases and support error-free training on GPU without accuracy loss originating from the mismatch between training and deployment. A PE is built based on this multiplier for parallel computation with 82% power reduction and 70% area reduction. Second, the proposed NMC-friendly data flow supports efficient data reuse and hence reduces off-chip memory traffic. The data flow supports data reuse of up to 16 times, matching the number of PEs and enabling regular read and write patterns. Third, a hardware-aware layer-wise model deployment approach is proposed with a memory space contiguity-aware (MSCA) model reshape strategy, and a hardware-aware NN splitting and scheduling algorithm. The proposed MSCA strategy maximizes burst access, and the proposed algorithm achieves efficient computation with high data reuse and low memory access. This deployment approach can achieve a reduction in memory access latency of 16.6%–32.0%. Measurements on a 22-nm test chip demonstrate a peak power efficiency of 33.98 TOPS/W under synthetic full-PE-utilization conditions, while achieving 12.92–29.11 TOPS/W for end-to-end NN inference on DarkNet19, VGG16, ViT-Tiny, and ResNet34.