3D Deformable Convolutions for MRI Classification

Pominova, Marina; Kondrateva, Ekaterina; Sharaev, Maxim; Bernstein, Alexander; Павлов, С. В.; Burnaev, Evgeny

doi:10.1109/icmla.2019.00278

Cited by 16 publications

(5 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One feasible approach is to capture the relationship between pixels within a reasonable area and suppress noise in the background. For this, we propose deformable convolution [35] to adaptively extract vessel features and strengthen the ability to refine thin vessels. This approach has achieved success in several tasks [36][37] [38][39], enabling receptive fields to automatically adapt to different sizes of vessels.…”

Section: ) Scale-aware Attention Modulementioning

confidence: 99%

Multi-Scale Deep Information and Adaptive Attention Mechanism Based Coronary Reconstruction of Superior Mesenteric Artery

et al. 2023

View full text Add to dashboard Cite

Vascular images contain a lot of key information, such as length, diameter and distribution. Thus reconstruction of vessels such as the Superior Mesenteric Artery is critical for the diagnosis of some abdominal diseases. However automatic segmentation of abdominal vessels is extremely challenging due to the multi-scale nature of vessels, boundary-blurring, low contrast, artifact disturbance and vascular cracks in Maximum Intensity Projection images. In this work, we propose a dual attention guided method where an adaptive adjustment field is applied to deal with multi-scale vessel information, and a channel feature fusion module is used to refine the extraction of thin vessels, reducing the interference and background noise. In particular, we propose a novel structure that accepts multiple sequential images as input, and successfully introduces spatial-temporal features by contextual information. A further IterUnet step is introduced to connect tiny cracks caused using CT scans. Comparing our proposed model with other state-of-the-art models, our model yields better segmentation and achieves an average F1 metric of 0.812.

show abstract

Section: ) Scale-aware Attention Modulementioning

confidence: 99%

Multi-Scale Deep Information and Adaptive Attention Mechanism Based Coronary Reconstruction of Superior Mesenteric Artery

et al. 2023

View full text Add to dashboard Cite

show abstract

“…For instance, the authors in [1] demonstrated that the prediction accuracy of the proposed DCN increases from 70% to 75% on the image semantic segmentation dataset (CityScapes). Significant prediction accuracy improvement is also observed in human motion recognition task [11] [12], action detection task [13] [14] and intelligent medical monitoring and treatment [15] [16].…”

Section: Introductionmentioning

confidence: 90%

Energy-Efficient Accelerator Design for Deformable Convolution Networks

Xu,

Chu,

Liu

et al. 2021

Preprint

View full text Add to dashboard Cite

Deformable convolution networks (DCNs) proposed to address the image recognition with geometric or photometric variations typically involve deformable convolution that convolves on arbitrary locations of input features. The locations change with different inputs and induce considerable dynamic and irregular memory accesses which cannot be handled by classic neural network accelerators (NNAs). Moreover, bilinear interpolation (BLI) operation that is required to obtain deformed features in DCNs also cannot be deployed on existing NNAs directly. Although a general purposed processor (GPP) seated along with classic NNAs can process the deformable convolution, the processing on GPP can be extremely slow due to the lack of parallel computing capability.To address the problem, we develop a DCN accelerator on existing NNAs to support both the standard convolution and deformable convolution. Specifically, for the dynamic and irregular accesses in DCNs, we have both the input and output features divided into tiles and build a tile dependency table (TDT) to track the irregular tile dependency at runtime. With the TDT, we further develop an on-chip tile scheduler to handle the dynamic and irregular accesses efficiently. In addition, we propose a novel mapping strategy to enable parallel BLI processing on NNAs and apply layer fusion techniques for more energy-efficient DCN processing. According to our experiments, the proposed accelerator achieves orders of magnitude higher performance and energy efficiency compared to the typical computing architectures including ARM, ARM+TPU, and GPU with 6.6% chip area penalty to a classic NNA.

show abstract

“…It has long been demonstrated that deformable 3D convolution has a large spatial receptive field. [15] Therefore, it models spatio-temporal information efficiently. In our work, we use five ResD3D blocks to model group-wise features F g n , as shown in Fig.…”

Section: B Spatio-temporal Alignmentmentioning

confidence: 99%

Grouped Spatio-Temporal Alignment Network for Video Super-Resolution

Zhang

2022

IEEE Signal Process. Lett.

View full text Add to dashboard Cite

Video super-resolution based on CNNs has recently achieved significant progress. The existing popular superresolution methods usually impose the optical flow between neighboring frames for temporal alignment. However, estimation of accurate optical flow is hard and expends greater computation.To address this problem, we propose a novel grouped spatiotemporal alignment network (GSTAN) that effectively incorporates spatio-temporal information in a hierarchical way. The input sequence is divided into several groups corresponding to different frame rates. These groups provide complementary information, which is helpful to restore missing textures for the reference frame. Specifically, each group employs deformable 3D convolution to incorporate spatio-temporal information, which avoids artifacts from explicit motion estimation. In addition, a Gated-Dconv information filter is proposed to control information flow to focus on the fine details. Finally, these groups provide complementary information, which is integrated with the intergroup fusion module. Extensive experiments have demonstrated our method achieves state-of-the-art SR performance on several benchmark datasets.

show abstract

3D Deformable Convolutions for MRI Classification

Cited by 16 publications

References 28 publications

Multi-Scale Deep Information and Adaptive Attention Mechanism Based Coronary Reconstruction of Superior Mesenteric Artery

Multi-Scale Deep Information and Adaptive Attention Mechanism Based Coronary Reconstruction of Superior Mesenteric Artery

Energy-Efficient Accelerator Design for Deformable Convolution Networks

Grouped Spatio-Temporal Alignment Network for Video Super-Resolution

Contact Info

Product

Resources

About