From Cloudiness to Clarity: Predictive Modeling for Early Diagnosis of Peritonitis in Peritoneal Dialysis (CLEAR-PD)
Project details
- Project number
- lzp-2025/1-0413
- Status
- Ongoing
- Period of time
- 01.01.2026 – 31.12.2028
- Financing
- 300,000
- Partners
- 3
- Coordinator
- Rīga Stradiņš university (RSU)
- Countries
- Latvia
- Project website
- https://www.rsu.lv/en/project/clear-pd(opens in a new window)
Project objectives and tasks
CLEAR-PD advances the early detection and management of peritoneal dialysis (PD)-associated peritonitis by developing accessible, automated diagnostic tools and targeted therapeutic alternatives. Traditional detection relies on subjective visual methods like the “newspaper test,” where delays directly increase the risk of catheter failure, hospitalization, and mortality. To overcome these limitations, CLEAR-PD combines physical and biochemical testing with machine learning models trained on smartphone or cost-efficient camera-captured dialysate images to reliably predict infection onset and severity at the point of care.
In parallel, the research evaluates the use of bacteriophages as both rapid diagnostic indicators and targeted therapies to eradicate biofilm-forming, antibiotic-resistant pathogens such as Staphylococcus aureus. By replacing subjective observations with data-driven monitoring and novel anti-biofilm solutions, the project seeks to enable timely clinical intervention, prevent complications, and improve outcomes for kidney disease patients.
EDI's tasks and role in the project
EDI is responsible for the machine learning and computer vision framework of CLEAR-PD. EDI's role focuses on developing a hybrid diagnostic system that integrates classical machine learning models - such as Support Vector Machines and Random Forests trained on biochemical test profiles - with convolutional neural networks (CNNs) designed to assess dialysate cloudiness from image data.
By engineering lightweight architectures suitable for deployment on consumer-grade hardware, EDI designs an efficient, point-of-care tool capable of providing automated risk predictions on low-compute devices, enabling integration into subsequent clinical testing and daily patient care.
This research is funded by the Latvian Council of Science, project CLEAR-PD, project No. lzp-2025/1-0413

