Development of the vascular condition classifier using supervised machine learning methods
- 27 November 2025
- journal article
- Published by Pirogov Russian National Research Medical University in Bulletin of Russian State Medical University
- No. 6
Abstract
Quantitative processing of optoacoustic angiograms is an important task, the solution of which will potentially enable the early diagnosis of vascular diseases. The aim of this study is to refine and conduct biomedical validation of a software tool for the analysis of optoacoustic angiograms, focusing on the application of machine learning methods. The work was conducted on an open dataset containing three-dimensional optoacoustic angiograms of an experimental animal (mouse) in three temperature conditions: cold temperature (16 °C), room temperature (23 °C), and body temperature (30 °C), as well as a dataset with basic vascular features obtained by processing using Amira/Avizo (Thermo Fisher Scientific), a general-purpose software for visualization and analysis of scientific and industrial data. Various vascular features missing from previous work were developed and calculated, after which basic methods of unsupervised/supervised clustering and supervised classification were applied to determine different temperature conditions of vessel segments. Supervised classification methods demonstrated high overall accuracy: CatBoost — 98.9%, SGDClassifier — 95.7%, and logistic regression — 99.7%. The results are consistent with existing descriptions of vascular changes during temperature tests. The applied methodology is universal, meaning with minor modifications it can be adapted to patients. Therefore, the results of this study may potentially improve the diagnosis of vascular pathologies.Keywords
This publication has 22 references indexed in Scilit:
- Photoacoustic Tomography in Cardiovascular Medicine: Innovations in Assessing Hemodynamics and Metabolic FunctionCurrent Treatment Options in Cardiovascular Medicine, 2025
- SKYQUANT 3D: Quantifying Vascular Anatomy With an Open‐Source Workflow for Comprehensive Analysis of Volumetric Optoacoustic Angiography DataJournal of Biophotonics, 2024
- Scanning optoacoustic angiography for assessing structural and functional alterations in superficial vasculature of patients with post-thrombotic syndrome: A pilot studyPhotoacoustics, 2024
- Dermal features derived from optoacoustic tomograms via machine learning correlate microangiopathy phenotypes with diabetes stageNature Biomedical Engineering, 2023
- Hybrid Photoacoustic Ultrasound Imaging System for Cold-Induced Vasoconstriction and Vasodilation MonitoringIEEE Transactions on Biomedical Engineering, 2023
- Quantification of vascular networks in photoacoustic mesoscopyPhotoacoustics, 2022
- The emerging role of photoacoustic imaging in clinical oncologyNature Reviews Clinical Oncology, 2022
- Seeing through the Skin: Photoacoustic Tomography of Skin Vasculature and BeyondJID Innovations, 2021
- Deep learning optoacoustic tomography with sparse dataNature Machine Intelligence, 2019
- Cardiovascular optoacoustics: From mice to men – A reviewPhotoacoustics, 2019