Work

Research on the application value of blood plasma autofluorescence spectroscopy model constructed by machine learning algorithms in identifying the types of post-operative kidney transplantation injury

Year: 2026

Type: article

Abstract: • Addresses a Critical Clinical Need: This study targets the urgent need for a minimally invasive, rapid diagnostic tool to differentiate complex post-renal transplant injury types (e.g., antibody-mediated rejection group, IgA nephropathy recurrence), potentially reducing reliance on invasive renal biopsies. • Introduces a Novel Methodological Approach: It is the first to apply blood plasma autofluorescence spectroscopy combined with machine learning for the specific classification of post-transplantation renal injury, presenting a new technological pathway in transplant monitoring. • Demonstrates Superior Diagnostic Performance: The model using blood plasma autofluorescence features alone achieved 84% accuracy in five-type classification, significantly outperforming traditional renal function indicators (36% accuracy). • Shows Exceptional Discriminatory Power for Stable vs. Unstable Grafts: In the binary classification (stable vs. unstable), the combined model achieved perfect discrimination (100% accuracy), highlighting its potential for initial patient triage. • Highlights Clinical Translation Potential: The technique is label-free, fast and based on routine blood samples, underscoring its strong potential for development into a practical clinical auxiliary diagnostic tool. This study aimed to assess the value of blood plasma autofluorescence spectroscopy in classifying renal injury after renal transplantation. The researchers included 86 patients after kidney transplantation and established five classifications: stable group, tubular injury group, IgA nephropathy group, antibody-mediated rejection (ABMR) group, and mixed group. They also established two classification (stable and unstable) diagnostic models based on renal histopathological results. Clinical and laboratory test data of patients were integrated, peripheral blood was collected and blood plasma autofluorescence spectra were detected; postoperative injury classification models were constructed using principal component analysis (PCA) and support vector machine (SVM). Among the five-classification diagnostic models, the overall accuracy of the classification obtained after training the model with blood plasma autofluorescence spectra alone was 84%; the overall accuracy of the classification obtained after training the model with laboratory renal function indexes alone was only 36%; and the accuracy of the classification obtained after combining the two with blood plasma was 88%. In the dichotomous diagnostic model, blood plasma autofluorescence spectroscopy alone had an overall accuracy of 96.30%; laboratory renal function alone had an overall accuracy of 84%; and the combination of both had an accuracy of 100%. Blood plasma autofluorescence spectroscopy has potential application in rapid, minimally invasive diagnosis of post-transplantation renal injury and identification of the type of injury.

Source: Photodiagnosis and Photodynamic Therapy

DOI: 10.1016/j.pdpdt.2026.105394

Authors: Lijuan Wu, Tangyuheng Liu, Wenjing Zhou, Wu Peng, Jianhua Zhu

Institutions: Sichuan University, West China Hospital of Sichuan University

Language: English

Topic: Renal Transplantation Outcomes and Treatments

Subfield: Transplantation

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $2,610