基于稀疏编码非局部注意力对偶网络的病理图像超分辨率重建  

Pathological Image Super-resolution Reconstruction Based on Sparse Coding Non-local Attention Dual Network

作　　者：梁美彦 张宇[1] 梁建安 陈庆辉 王茹 王琳 LIANG Meiyan;ZHANG Yu;LIANG Jianan;CHEN Qinghui;WANG Ru;WANG Lin(College of Physics and Electronics Engineering,Shanxi University,Taiyuan 030006,China;Shanxi Bethune Hospital,Third Hospital of Shanxi Medical University,Taiyuan 030032,China;Tongji Hospital Tongji Medical College,Huazhong University of Science and Technology,Wuhan 430030,China)

机构地区：[1]山西大学物理电子工程学院,太原030006 [2]山西白求恩医院,山西医科大学第三医院,太原030032 [3]华中科技大学同济医学院附属同济医院,武汉430030

出　　处：《计算机科学》2023年第S01期295-302,共8页Computer Science

基　　金：国家自然科学基金(11804209);山西省面上青年基金(201901D211173);山西省高校科技创新计划(2019L0064);山西省应用基础研究计划项目(201901D211172);山西省中央引导地方科技发展资金项目(YDZX20201400001547);山西省高等学校科技创新项目(2020L0048)。

摘　　要：高分辨率的病理学图像是疾病高精度诊断的客观依据,在精准医学领域具有重要意义。然而,受硬件设备分辨率和扫描时长的限制,实时获取高分辨率病理图像存在困难。经典的图像超分辨率重建算法由于模型的参数较难估计,导致重建后图像细节模糊且不够真实,不适用于病理学图像。为此,文中提出稀疏编码非局部注意力对偶网络,通过上采样和降采样对偶分支中的稀疏编码非局部注意力机制、高斯约束以及参数共享策略来实现病理学图像的超分辨率重建。重建后的病理图像峰值信噪比和结构相似性分别达到了30.84 dB和0.914。研究结果表明,所提方法不但能够实现病理学图像中高频细节的精确重建,轻量化的稀疏编码非局部注意力机制也有效地提高了建模的效率,是病理学图像超分辨率重建的一种有效方法。High-resolution pathological images are the objective criteria for high-precision disease diagnosis,which have great significance in the field of precision medicine.However,it is difficult to obtain high-resolution pathological images in real time,due to the limited resolution and constrained scanning time of hardware devices.Classical image super-resolution reconstruction algorithm is not suitable for pathological images because the parameters of the model are difficult to estimate,resulting in blurred and unrealistic image details after super-resolution reconstruction.Therefore,sparse-coding non-local attention dual network(SNADN)is proposed,which uses Gaussian constraints,hash coding and parameter sharing strategy in the dual branches,to achieve high accuracy and high efficiency super-resolution reconstruction of pathological images.The PSNR and SSIM of the reconstructed pathological images can reach 30.84 dB and 0.914,respectively.The results show that SNADN can not only achieve accurate reconstruction of high-frequency details in pathological images,but also the lightweight sparse coding non-local attention mechanism can effectively improve the modeling efficiency.It is an effective method for super-resolution reconstruction of pathological images.

关 键 词：稀疏编码 非局部注意力 对偶网络 病理图像 超分辨率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]