TY - JOUR

T1 - Approach to non-invasive screening of kidney and prostate cancer via potentiometric multisensor urine analysis and machine learning

AU - Юськина, Екатерина Андреевна

AU - Мосоян, Михаил

AU - Джагацпанян, Игорь Эдуардович

AU - Васильев , Артем

AU - Макеев, Владимир

AU - Гапонова, Анна Георгиевна

AU - Протощак, Владимир Владимирович

AU - Карпущенко, Евгений Геннадьевич

AU - Слепцов, Александр

AU - Кирсанов, Дмитрий Олегович

PY - 2025/11

Y1 - 2025/11

N2 - The chemical composition of biofluids (serum, urine) for patients with a diagnosed cancer disease and healthy people can differ significantly. This study reports on the development of a simple potentiometric multisensor system capable for distinguishing urine samples from prostate cancer patients, kidney cancer patients and healthy controls. The sensors respond to the general ionic composition of the samples and provide characteristic patterns for further processing. 116 urine samples were analyzed: 39 from patients with diagnosed prostate cancer, 38 from patients with diagnosed kidney cancer and 39 from control group. Potentiometric sensor responses have been used as input data for various machine learning methods: logistic regression (LR), k-nearest neighbors (kNN), XGBoost classifier (XGBC), random forest (RF), and support vector machines (SVM). The SVM classifier demonstrates the highest classification accuracy of 77 % when distinguishing between urine samples from kidney cancer patients/ control group, and accuracy 79 % when distinguishing between prostate cancer/ control group. Moreover, the potentiometric multisensor system can be used to distinguish samples from patients with different oncological diseases such as kidney cancer and prostate cancer with an accuracy of 87 % for RF classification method. Considering the non-invasiveness of the approach proposed the method being properly validated with external data on the extended number of samples may become a promising tool for screening kidney cancer and prostate cancer at the same time. Once the suspicious samples are identified with potentiometric system they can be further validated through the standard clinical diagnostic protocols.

AB - The chemical composition of biofluids (serum, urine) for patients with a diagnosed cancer disease and healthy people can differ significantly. This study reports on the development of a simple potentiometric multisensor system capable for distinguishing urine samples from prostate cancer patients, kidney cancer patients and healthy controls. The sensors respond to the general ionic composition of the samples and provide characteristic patterns for further processing. 116 urine samples were analyzed: 39 from patients with diagnosed prostate cancer, 38 from patients with diagnosed kidney cancer and 39 from control group. Potentiometric sensor responses have been used as input data for various machine learning methods: logistic regression (LR), k-nearest neighbors (kNN), XGBoost classifier (XGBC), random forest (RF), and support vector machines (SVM). The SVM classifier demonstrates the highest classification accuracy of 77 % when distinguishing between urine samples from kidney cancer patients/ control group, and accuracy 79 % when distinguishing between prostate cancer/ control group. Moreover, the potentiometric multisensor system can be used to distinguish samples from patients with different oncological diseases such as kidney cancer and prostate cancer with an accuracy of 87 % for RF classification method. Considering the non-invasiveness of the approach proposed the method being properly validated with external data on the extended number of samples may become a promising tool for screening kidney cancer and prostate cancer at the same time. Once the suspicious samples are identified with potentiometric system they can be further validated through the standard clinical diagnostic protocols.

KW - Kidney cancer

KW - Machine learning

KW - Multisensor system

KW - Non-invasive screening

KW - Potentiometry

KW - Prostate cancer

KW - Urine analysis

UR - https://www.mendeley.com/catalogue/4dd282a8-414d-3111-a8b5-c0992b09eacc/

U2 - 10.1016/j.microc.2025.115589

DO - 10.1016/j.microc.2025.115589

M3 - Article

VL - 218

JO - Microchemical Journal

JF - Microchemical Journal

SN - 0026-265X

M1 - 115589

ER -