Ding Huanfei, Fawad Muhammad, Xu Xiaolin, Hu Bowen
The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China.
School of Public Health and Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China.
Front Oncol. 2022 Oct 14;12:1048348. doi: 10.3389/fonc.2022.1048348. eCollection 2022.
Hepatocellular carcinoma (HCC) is one of the most commonly seen liver disease. Most of HCC patients are diagnosed as Hepatitis B related cirrhosis simultaneously, especially in Asian countries. HCC is the fifth most common cancer and the second most common cause of cancer-related death in the World. HCC incidence rates have been rising in the past 3 decades, and it is expected to be doubled by 2030, if there is no effective means for its early diagnosis and management. The improvement of patient's care, research, and policy is significantly based on accurate medical diagnosis, especially for malignant tumor patients. However, sometimes it is really difficult to get access to advanced and expensive diagnostic tools such as computed tomography (CT), magnetic resonance imaging (MRI) and positron emission tomography (PET-CT)., especially for people who resides in poverty-stricken area. Therefore, experts are searching for a framework for predicting of early liver diseases based on basic and simple examinations such as biochemical and routine blood tests, which are easily accessible all around the World. Disease identification and classification has been significantly enhanced by using artificial intelligence (AI) and machine learning (ML) in conjunction with clinical data. The goal of this research is to extract the most significant risk factors or clinical parameters for liver diseases in 525 patients based on clinical experience using machine learning algorithms, such as regularized regression (RR), logistic regression (LR), random forest (RF), decision tree (DT), and extreme gradient boosting (XGBoost). The results showed that RF classier had the best performance (accuracy = 0.762, recall = 0.843, F1-score = 0.775, and AUC = 0.999) among the five ML algorithms. And the important orders of 14 significant risk factors are as follows: Total bilirubin, gamma-glutamyl transferase (GGT), direct bilirubin, hemoglobin, age, platelet, alkaline phosphatase (ALP), aspartate transaminase (AST), creatinine, alanine aminotransferase (ALT), cholesterol, albumin, urea nitrogen, and white blood cells. ML classifiers might aid medical organizations in the early detection and classification of liver disease, which would be beneficial in low-income regions, and the relevance of risk factors would be helpful in the prevention and treatment of liver disease patients.
肝细胞癌（HCC）是最常见的肝脏疾病之一。大多数HCC患者同时被诊断为乙型肝炎相关肝硬化，尤其是在亚洲国家。HCC是全球第五大常见癌症，也是癌症相关死亡的第二大常见原因。在过去30年中，HCC发病率一直在上升，如果没有有效的早期诊断和管理手段，预计到2030年将翻倍。患者护理、研究和政策的改善很大程度上基于准确的医学诊断，尤其是对于恶性肿瘤患者。然而，有时很难获得先进且昂贵的诊断工具，如计算机断层扫描（CT）、磁共振成像（MRI）和正电子发射断层扫描（PET-CT），特别是对于居住在贫困地区的人。因此，专家们正在寻找一种基于生化和血常规等基础且简单检查来预测早期肝脏疾病的框架，这些检查在世界各地都很容易获得。通过将人工智能（AI）和机器学习（ML）与临床数据相结合，疾病识别和分类得到了显著增强。本研究的目的是基于临床经验，使用机器学习算法，如正则化回归（RR）、逻辑回归（LR）、随机森林（RF）、决策树（DT）和极端梯度提升（XGBoost），从525例患者中提取肝脏疾病最重要的危险因素或临床参数。结果表明，在五种ML算法中，RF分类器表现最佳（准确率=0.762，召回率=0.843，F1分数=0.775，AUC=0.999）。14个重要危险因素的重要性排序如下：总胆红素、γ-谷氨酰转移酶（GGT）、直接胆红素、血红蛋白、年龄、血小板、碱性磷酸酶（ALP）、天冬氨酸转氨酶（AST）、肌酐、丙氨酸转氨酶（ALT）、胆固醇、白蛋白、尿素氮和白细胞。ML分类器可能有助于医疗机构对肝脏疾病进行早期检测和分类，这对低收入地区有益，危险因素的相关性将有助于肝病患者的预防和治疗。
