Chen Sheng, Cai Yuantao
Department of automatic control, school of optical-electrical and computer engineering, University of Shanghai for Science and Technology, 516 Jungong Road, Shanghai, 200093, China.
Diagnostics (Basel). 2019 Apr 24;9(2):45. doi: 10.3390/diagnostics9020045.
In intensive care units (ICUs), supporting devices play an important role, and the placement of these devices must be accurate, such as catheters and tubes. Taking portable chest radiograph (CXRs) for patients in ICU is a standard procedure. However, non-optimized exposure settings and misaligned body positions usually mean that portable CXRs are not in acceptable working condition. The purpose of this study was to enhance ICU CXRs to assist radiologists in the positioning of endotracheal, feeding, and nasogastric tubes in ICU patients. The unsharp masking model (USM) was a classical image enhancement technique. Because of the isotropic diffusion filter applied in this model, USM enhanced the edge information and noise simultaneously. In this paper, we proposed a reverse anisotropic diffusion (RAD)-based USM technique for enhancement of line structures in ICU CXRs. First, a RAD algorithm was applied to replace the Gaussian filter in the classical USM. The RAD algorithm only produced a smoothed image, in which edge information was smoothed while the noise was preserved. Then, the smoothed image was subtracted from the original image to produce the unsharp mask whereby only the edges were retained. Consequently, only edge information was enhanced in the final enhanced image by using the RAD-based USM model. The proposed method was tested for 87 ICU CXRs and the findings indicate that this approach can enhance image edges efficiently while suppressing noise.
在重症监护病房（ICU）中，支撑设备起着重要作用，这些设备的放置必须准确无误，例如导管等。为ICU患者拍摄便携式胸部X光片（CXR）是一项标准程序。然而，曝光设置未优化以及身体位置未对齐通常意味着便携式CXR的成像效果不佳。本研究的目的是改进ICU的CXR图像，以协助放射科医生对ICU患者的气管内导管、饲管和鼻胃管进行定位。非锐化掩模模型（USM）是一种经典的图像增强技术。由于该模型应用了各向同性扩散滤波器，USM在增强边缘信息的同时也增强了噪声。在本文中，我们提出了一种基于反向各向异性扩散（RAD）的USM技术，用于增强ICU的CXR图像中的线条结构。首先，应用RAD算法替代经典USM中的高斯滤波器。RAD算法仅生成一个平滑后的图像，其中边缘信息被平滑处理，而噪声得以保留。然后，从原始图像中减去平滑后的图像以生成非锐化掩模，从而仅保留边缘部分。因此，通过使用基于RAD的USM模型，最终增强后的图像中仅边缘信息得到了增强。该方法在87张ICU的CXR图像上进行了测试，结果表明该方法能够在抑制噪声的同时有效地增强图像边缘。
