Xiao Yiming, Boily Mathieu, Hashemi Hoda Sadat, Rivaz Hassan
PERFORM Centre, Concordia University, Montreal, Quebec, Canada; Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, Canada.
Department of Diagnostic Radiology, McGill University, Montreal, Quebec, Canada.
Ultrasound Med Biol. 2018 Jul;44(7):1525-1532. doi: 10.1016/j.ultrasmedbio.2018.03.004. Epub 2018 Apr 5.
Raw ultrasound (US) signal has a very high dynamic range (HDR) and, as such, is compressed in B-mode US using a logarithmic function to fit within the dynamic range of digital displays. However, in some cases, hyper-echogenic tissue can be overexposed at high gain levels with the loss of hypo-echogenic detail at low gain levels. This can cause the loss of anatomic detail and tissue texture and frequent and inconvenient gain adjustments, potentially affecting the diagnosis. To mitigate these drawbacks, we employed tone mapping operators (TMOs) in HDR photography to create HDR US. We compared HDR US produced from three different popular TMOs (Reinhard, Drago and Durand) against conventional US using a simulated US phantom and in vivo images of patellar tendon pathologies. Based on visual inspection and assessments of structural fidelity, image entropy and contrast-to-noise ratio metrics, Reinhard and Drago TMOs substantially improved image detail and texture.
原始超声（US）信号具有非常高的动态范围（HDR），因此，在B模式超声中使用对数函数进行压缩，以适应数字显示器的动态范围。然而，在某些情况下，高回声组织在高增益水平下可能会过度曝光，而低回声细节在低增益水平下会丢失。这可能会导致解剖细节和组织纹理的丢失，以及频繁且不方便的增益调整，从而可能影响诊断。为了减轻这些缺点，我们在HDR摄影中采用色调映射算子（TMO）来创建HDR US。我们使用模拟US体模和髌腱病变的体内图像，将三种不同的流行TMO（Reinhard、Drago和Durand）生成的HDR US与传统US进行了比较。基于目视检查以及对结构保真度、图像熵和对比度噪声比指标的评估，Reinhard和Drago TMO显著改善了图像细节和纹理。
