We proposed a novel self-supervised learning pretext task, involving three transformations— cube ordering, rotating and masking, to pre-train 3D neural networks for volumetric medical images
GLSVLSI '20: Great Lakes Symposium on VLSI 2020
Virtual Event
China
September,..., pp.457-462, (2020)
Processing-in-Memory (PIM) or Near-Data Processing has been recognized as the most potential solution to resolve the ever-aggravating memory wall especially as the thrive of memory-intensive scale-out workloads such as graph computing and data analytics. However, when the future ...
medical image computing and computer assisted intervention, pp.417-428, (2020)
(ULD) in computed tomography plays an essential role in computer-aided diagnosis systems. Many detection approaches achieve excellent results for ULD using possible bounding boxes (or anchors) as proposals. However, empirical evidence shows that using anchor-based proposals leads...
medical image computing and computer-assisted intervention, pp.692-702, (2020)
Recent methods in multiple landmark detection based on deep convolutional neural networks (CNNs) reach high accuracy and improve traditional clinical workflow. However, the vulnerability of CNNs to adversarial-example attacks can be easily exploited to break classification and se...
Carbon Nanotubu field-effect transistor (CNFET) that promises both higher clock speed and energy efficiency becomes an attractive alternative to the conventional power-hungry CMOS cache. We observe that the CNFET-based cache constructed with typical SRAM cells has distinct energy...
11Xeon Gold 6154 can only support multiplication between equal bitwidth fixed-point numbers, so in this experiment int16 × int8 is implemented as int16 × int16
Chen Weiwei,Wang Ying,Yang Shuang,Liu Chen,Zhang Lei
DATE, pp.1283-1286, (2020)
DNN/Accelerator co-design has shown great potential in improving QoR and performance. Typical approaches separate the design flow into two-stage: (1) designing an application-specific DNN model with high accuracy; (2) building an accelerator considering the DNN specific charact...
We conduct experiments and clinical evaluations based on two benchmarking Chest X-rays databases and demonstrate that the performances of real clinical diagnosis and automatic lung disease classification are boosted with the aid of the bone suppressed image
When using atom-centered integration grids, the portion of the grid that belongs to a certain atom also moves when this atom is displaced. In the paper, we investigate the moving-grid effect in the calculation of the harmonic vibrational frequencies when using all-electron full-p...
IEEE transactions on medical imaging, no. 3 (2019): 634-643
Current deep neural network based approaches to computed tomography (CT) metal artifact reduction (MAR) are supervised methods that rely on synthesized metal artifacts for training. However, as synthesized data may not accurately simulate the underlying physical mechanisms of CT ...
The significant gap between the unique feature of graph processing and the hardware features of general-purpose architectures limits the further improvement of performance and energy efficiency
Proceedings of the International Conference for High Performance Computing, Networking, Storage and ..., pp.68, (2019)
With more attention attached to nuclear energy, the formation mechanism of the solute clusters precipitation within complex alloys becomes intriguing research in the embrittlement of nuclear reactor pressure vessel (RPV) steels. Such phenomenon can be simulated with atomic kineti...
It is noteworthy that while smart-retransmission time out seems more aggressive than ER and Tail Loss Probe, the retransmission rate is still kept small as we show through experiments