Ramkumar M, Kalirajan K, Kumar U Pavan, Surya P
Sri Krishna College of Engineering and Technology, Coimbatore, Tamil Nadu 641008 India.
KPR Institute of Engineering and Technology, Coimbatore, Tamil Nadu 641407 India.
Biomed Eng Lett. 2023 Oct 31;14(2):221-233. doi: 10.1007/s13534-023-00332-5. eCollection 2024 Mar.
Lymph node metastasis detections are more clinically significant task associated with the presence and reappearance of lung cancer. The development of the computer-assisted diagnostic approach has greatly supported the diagnosis of human disorders in the field of medicine including lung cancer. Lung cancer treatment is possible if it is detected at the initial stage. Radiologists have great difficulty identifying and categorizing lung cancers in the initial phase. So, several methods were used to predict the lung cancer but does not provide accurate solutions with increased error rate. To overcome these issues, a Deep Volcanic Residual U-Net (DVR U-Net) for nodal metastasis is proposed in this manuscript which identifies the LC accurately in the early stage. Initially, the input images are taken from two datasets. After that, these input data are pre-processed using Anisotropic Diffusion Filter with a Fuzzy based Contrast-Limited Adaptive Histogram Equalization (ADFFCLAHE) method Then the pre-processed images are given to the DVR U-Net to segment and extract the volume of interest for estimating the nodal stage of each volume of interest. Finally, DVR U-Net effectively detects and classifies the N + (nodal metastasis) or N- (non-nodal metastasis). The introduced method attains 99.9% higher accuracy as compared with the existing methods. Also, the statistical analysis of the Shapiro-Wilk test, Friedman test and Wilcoxon Signed-Rank test are executed to prove the statistical effectiveness of the implemented method.
淋巴结转移检测是与肺癌的存在和复发相关的更具临床意义的任务。计算机辅助诊断方法的发展极大地支持了包括肺癌在内的医学领域人类疾病的诊断。如果肺癌在初始阶段被检测到，那么治疗是有可能的。放射科医生在肺癌初始阶段的识别和分类方面存在很大困难。因此，人们使用了几种方法来预测肺癌，但这些方法不能提供准确的解决方案，错误率还在增加。为了克服这些问题，本文提出了一种用于淋巴结转移的深度火山残差U-Net（DVR U-Net），它能在早期准确识别肺癌。首先，输入图像取自两个数据集。之后，使用基于模糊的对比度受限自适应直方图均衡化的各向异性扩散滤波器（ADFFCLAHE）方法对这些输入数据进行预处理。然后将预处理后的图像输入到DVR U-Net中，以分割并提取感兴趣区域的体积，用于估计每个感兴趣区域的淋巴结分期。最后，DVR U-Net有效地检测并分类N +（淋巴结转移）或N -（无淋巴结转移）。与现有方法相比，所提出的方法准确率提高了99.9%。此外，还进行了夏皮罗-威尔克检验、弗里德曼检验和威尔科克森符号秩检验的统计分析，以证明所实施方法的统计有效性。
