Intelligence recognition of clotting level in hemodialysis

The consequences of clotting formation during hemodialysis are highly consequential, highlighting the need for a detection technology that offers real-time responsiveness and high accuracy to monitor in context hemodialysis. Currently, primarily relies on visual observation by medical workers offline testing following scheduled blood sampling, which falls short providing insights into status. In this study, we present device based MS-UNet++ framework, capable identifying areas assessing levels real time through venous pot images within circuit. Clinical experiments have demonstrated model achieves an 84.0% classifying levels. Furthermore, when it comes alerting exceeding threshold, exhibits sensitivity 100%. These results affirm ability framework accurately identify process provide timely warnings necessary. stability applicability neural network been verified, offering novel approaches monitoring