Siri Sangeeta K, Latte Mrityunjaya V
Department of Electronics & Communication Engineering, Sapthagiri College of Engineering, Bengaluru, karnataka 560057, India.
JSS Academy of Technical Education, Bengaluru, India.
Comput Methods Programs Biomed. 2017 Nov;151:101-109. doi: 10.1016/j.cmpb.2017.08.020. Epub 2017 Aug 24.
Many different diseases can occur in the liver, including infections such as hepatitis, cirrhosis, cancer and over effect of medication or toxins. The foremost stage for computer-aided diagnosis of liver is the identification of liver region. Liver segmentation algorithms extract liver image from scan images which helps in virtual surgery simulation, speedup the diagnosis, accurate investigation and surgery planning. The existing liver segmentation algorithms try to extort exact liver image from abdominal Computed Tomography (CT) scan images. It is an open problem because of ambiguous boundaries, large variation in intensity distribution, variability of liver geometry from patient to patient and presence of noise. A novel approach is proposed to meet challenges in extracting the exact liver image from abdominal CT scan images. The proposed approach consists of three phases: (1) Pre-processing (2) CT scan image transformation to Neutrosophic Set (NS) and (3) Post-processing. In pre-processing, the noise is removed by median filter. The "new structure" is designed to transform a CT scan image into neutrosophic domain which is expressed using three membership subset: True subset (T), False subset (F) and Indeterminacy subset (I). This transform approximately extracts the liver image structure. In post processing phase, morphological operation is performed on indeterminacy subset (I) and apply Chan-Vese (C-V) model with detection of initial contour within liver without user intervention. This resulted in liver boundary identification with high accuracy. Experiments show that, the proposed method is effective, robust and comparable with existing algorithm for liver segmentation of CT scan images.
肝脏会出现许多不同的疾病，包括肝炎等感染性疾病、肝硬化、癌症以及药物或毒素的过量影响。肝脏计算机辅助诊断的首要阶段是肝脏区域的识别。肝脏分割算法从扫描图像中提取肝脏图像，这有助于虚拟手术模拟、加快诊断速度、进行准确的检查和手术规划。现有的肝脏分割算法试图从腹部计算机断层扫描（CT）图像中精确提取肝脏图像。由于边界模糊、强度分布差异大、不同患者肝脏几何形状的变异性以及噪声的存在，这是一个开放性问题。本文提出了一种新颖的方法来应对从腹部CT扫描图像中提取精确肝脏图像的挑战。所提出的方法包括三个阶段：（1）预处理；（2）将CT扫描图像转换为中智集（NS）；（3）后处理。在预处理阶段，通过中值滤波器去除噪声。设计“新结构”将CT扫描图像转换到中智域，该中智域用三个隶属子集表示：真子集（T）、假子集（F）和不确定子集（I）。这种转换大致提取了肝脏图像结构。在后处理阶段，对不确定子集（I）进行形态学操作，并应用Chan-Vese（C-V）模型在无需用户干预的情况下检测肝脏内的初始轮廓。这使得能够高精度地识别肝脏边界。实验表明，所提出的方法对于CT扫描图像的肝脏分割是有效、稳健的，并且与现有算法具有可比性。
