Hanzhi Chen,1,* Leting Zhou,1,* Meilin Yan,1,* Cheng Li,1 Bin Liu,1 Xiaobin Liu,1 Weiwei Shan,1 Ya Guo,2 Zhijian Zhang,1 Liang Wang1 

1Department of Nephrology, The Affiliated Wuxi People’s Hospital of Nanjing Medical University, Wuxi People’s Hospital, Wuxi Medical Center, Nanjing Medical University, Wuxi, Jiangsu, 214000, People’s Republic of China; 2Key Laboratory of Advanced Process Control for Light Industry (Ministry of Education), Jiangnan University, Wuxi, Jiangsu, 214122, People’s Republic of China

*These authors contributed equally to this work

Correspondence: Zhijian Zhang, The Affiliated Wuxi People’s Hospital of Nanjing Medical University, Wuxi People’s Hospital, Wuxi Medical Center, Nanjing Medical University, 299 Qingyang Road, Wuxi, Jiangsu, 214000, People’s Republic of China, Email zzjwxyy@163.com Liang Wang, The Affiliated Wuxi People’s Hospital of Nanjing Medical University, 299 Qingyang Road, Wuxi, Jiangsu, 214000, People’s Republic of China, Email wangliang_wuxi@126.com

Background: End-stage kidney disease (ESKD) patients often face complications like anemia, malnutrition, and cardiovascular issues. Serological tests, which are uncomfortable and not frequently conducted, assist in medical assessments. A non-invasive, convenient method for determining these test results would be beneficial for monitoring patient health.
Objective: This study develops machine learning models to estimate key serological test results using non-invasive cellular bioelectrical impedance measurements, a routine procedure for ESKD patients.
Methods: The study employs two machine learning models, Support Vector Machine (SVM) and Random Forest (RF), to determine key serological tests by classifying cell bioelectrical indicators. Data from 688 patients, comprising 3,872 biochemical–bioelectrical records, were used for model validation.
Results: Both SVM and RF models effectively categorized key serological results (albumin, phosphorus, parathyroid hormone) into low, normal, and high. RF generally outperformed SVM, except in classifying calcium levels in women.
Conclusion: The machine learning models effectively classified serological test results for maintenance hemodialysis patients using cellular bioelectrical indicators, therefore can help in making judgments about physicochemical indicators using electrical signals, thereby reducing the frequency of serological tests.

Keywords: serological test results, cellular bioelectrical indicators, machine learning, End-stage kidney disease