Comprehensive Summary
Egen et al. investigates how a machine learning (ML) model could analyze the hyperspectral imaging (HSI) data and detect renal malperfusions caused by elevated intracranial pressure (IRP). The authors caused hydronephrosis within the porcine models by clamping the ureters at 30, 50, 70, and 90 mmHg and recorded 1,744 hyperspectral images from 73 pigs. The authors utilized the ML model to interpret the spectral data; the goal was to separate physiological perfusion from states of ischemia and venous congestion. The model classified most hydronephrotic kidneys as ischemic. This demonstrated the effectiveness of ML in detecting malperfusion and capturing the pathomechanisms of hydronephrosis. The study also used HyperGUI, a graphic interface tool that helps organize ML data, making the results more consistent. Together, these tools may allow for automated detection of malperfusion without relying solely on human interpretation.
Outcomes and Implications
The combination of ML and HyperGUI highlights how AI can simplify complex imaging data into clear and more accessible results. This approach has the potential of providing clinicians with real-time decision support during procedures. However, the findings are limited to an animal model, with no testing in different groups or settings, and safe pressure limits for human patients remain unproven. Despite these limitations, demonstrating ML model’s ability to classify porcine kidney perfusion states is encouraging.