Faculty of Biological Sciences, Quaid-i-Azam University, Islamabad, Pakistan.
Department of gastroenterology and hepatology, Pakistan Kidney and Liver Institute, Lahore, Pakistan.
Photodiagnosis Photodyn Ther. 2023 Sep;43:103677. doi: 10.1016/j.pdpdt.2023.103677. Epub 2023 Jun 29.
Current diagnostic methods for assessment of hepatitis C virus related hepatocellular carcinoma and subsequent categorization of hepatocellular carcinoma into non-angio-invasive hepatocellular carcinoma and angio-invasive hepatocellular carcinoma, to establish appropriate treatment strategies, are costly, invasive and requires multiple screening steps. This demands alternative diagnostic approaches that are cost-effective, time-efficient, and minimally invasive, while maintaining their efficacy for screening of hepatitis c virus related hepatocellular carcinoma. In this study, we propose that attenuated total reflection Fourier transform infrared in conjunction with principal component analysis - linear discriminant analysis and support vector machine multivariate algorithms holds a potential as a sensitive tool for the detection of hepatitis C virus-related hepatocellular carcinoma and the subsequent categorization of hepatocellular carcinoma into non-angio-invasive hepatocellular carcinoma and angio-invasive hepatocellular carcinoma.
Freeze-dried sera samples collected from 31 hepatitis c virus related hepatocellular carcinoma patients and 30 healthy individuals, were used to acquire mid-infrared absorbance spectra (3500-900 cm-) using attenuated total reflection Fourier transform infrared. Chemometric machine learning techniques were utilized to build principal component analysis - linear discriminant analysis and support vector machine discriminant models for the spectral data of hepatocellular carcinoma patients and healthy individuals. Sensitivity, specificity, and external validation on blind samples were calculated.
Major variations were observed in the two spectral regions i.e., 3500-2800 and 1800-900 cm-. IR spectral signatures of hepatocellular carcinoma were reliably different from healthy individuals. Principal component analysis - linear discriminant analysis and support vector machine models computed 100% accuracy for diagnosing hepatocellular carcinoma. To classify the non-angio-invasive hepatocellular carcinoma/ angio-invasive hepatocellular carcinoma status, diagnostic accuracy of 86.21% was achieved for principal component analysis - linear discriminant analysis. While the support vector machine showed a training accuracy of 98.28% and a cross-validation accuracy of 82.75%. External validation for support vector machine based classification observed 100% sensitivity and specificity for accurately classifying the freeze-dried sera samples for all categories.
We present the specific spectral signatures for non-angio-invasive hepatocellular carcinoma and angio-invasive hepatocellular carcinoma, which were prominently differentiated from healthy individuals. This study provides an initial insight into the potential of attenuated total reflection Fourier transform infrared to diagnose hepatitis C virus related hepatocellular carcinoma but also to further categorize into non-angio-invasive and angio-invasive hepatocellular carcinoma.
目前用于评估丙型肝炎病毒相关肝细胞癌的诊断方法以及将肝细胞癌分为非血管侵袭性肝细胞癌和血管侵袭性肝细胞癌的后续分类，以制定适当的治疗策略，其费用昂贵、具有侵入性并且需要多个筛查步骤。这就需要有替代的诊断方法，这些方法具有成本效益、耗时少且微创，同时保持对丙型肝炎病毒相关肝细胞癌筛查的有效性。在本研究中，我们提出衰减全反射傅里叶变换红外与主成分分析-线性判别分析和支持向量机多元算法相结合，有望成为一种敏感的工具，用于检测丙型肝炎病毒相关肝细胞癌，并将肝细胞癌进一步分为非血管侵袭性肝细胞癌和血管侵袭性肝细胞癌。
从 31 名丙型肝炎病毒相关肝细胞癌患者和 30 名健康个体采集冻干血清样本，使用衰减全反射傅里叶变换红外获取中红外吸收光谱（3500-900 cm-1）。利用化学计量机器学习技术为肝细胞癌患者和健康个体的光谱数据构建主成分分析-线性判别分析和支持向量机判别模型。计算了敏感性、特异性和盲样的外部验证。
在两个光谱区域即 3500-2800 和 1800-900 cm-1 观察到主要变化。肝细胞癌的红外光谱特征与健康个体可靠地区别开来。主成分分析-线性判别分析和支持向量机模型对诊断肝细胞癌的准确率达到 100%。为了对非血管侵袭性肝细胞癌/血管侵袭性肝细胞癌的状态进行分类，主成分分析-线性判别分析的诊断准确率为 86.21%。而支持向量机的训练准确率为 98.28%，交叉验证准确率为 82.75%。支持向量机分类的外部验证观察到，所有类别中对冻干血清样本的分类都达到了 100%的敏感性和特异性。
我们提出了非血管侵袭性肝细胞癌和血管侵袭性肝细胞癌的特定光谱特征，它们与健康个体明显不同。本研究初步探讨了衰减全反射傅里叶变换红外诊断丙型肝炎病毒相关肝细胞癌的潜力，也进一步将其分为非血管侵袭性和血管侵袭性肝细胞癌。
