Research On Classification Of Dialysis Pressure Curve Based On Markov Chain And Change-point Detection

Hemodialysis is one of the effective treatment methods for patients with acute and chronic renal failure,and the safety of hemodialysis is particularly important.How to use the changes in arterial and venous pressure,transmembrane pressure,and other pressure curves of hemodialysis machines to predict the critical point at which dialysis safety will transition to a critical state in advance has important medical research value.Starting from the perspective of curve change point detection and combining with Markov chain prediction,this article proposes a two-stage change point detection algorithm of "initial screening first,then testing" to classify dialysis pressure curves.Firstly,the transmembrane pressure curve of coagulation patients was selected as the research object,and ten state intervals were adaptively divided based on different quantiles of the patient’s own transmembrane pressure data.Establish a real-time rolling prediction mechanism for transmembrane pressure data using Markov chain and sliding window methods.Using mean square error as a predictive evaluation indicator,draw a change point intensity map and set a change point intensity threshold to achieve the initial screening process of the transmembrane pressure data curve.Secondly,with a fixed interval span of 25mm Hg,all transmembrane pressure data are uniformly divided into ten state intervals,and the transmembrane pressure state sequence is calculated based on the state of each pressure value.Traverse through all the initially screened change points one by one,and calculate the state transition frequency matrix before and after each change point.Taking the product of different quantile as the weight coefficient of the transfer between states,a weight matrix based on ten quantiles of the patient’s own transmembrane pressure data was constructed.After weighted processing,calculate the differential information matrix,and use the sum of the elements of the differential information matrix as the value of the differential information indicator.According to the difference information curve after one differential processing,the maximum peak is the sequence of the maximum change point that can truly reflect the global difference information.With this Sequence point,the original data point at the maximum change point is restored,and the "false" change points caused by pressure outlier are effectively eliminated,so as to realize the secondary detection of change points in the transmembrane pressure data curve.Finally,the traditional TMP-250mmHg single point detection method and the two-stage change point detection algorithm proposed in this article were applied to the example data of a hospital in Sichuan Province.The results showed that the two-stage change point detection algorithm based on this article had a good classification effect on the dialysis pressure curve and significantly improved the accuracy of coagulation classification.