Multilayer Perceptron-Based Real-Time Intradialytic Hypotension Prediction Using Patient Baseline Information and Heart-Rate Variation

Abstract

Intradialytic hypotension (IDH) is a common side effect that occurs during hemodialysis and poses a great risk for dialysis patients. Many studies have been conducted so far to predict IDH, but most of these could not be applied in real-time because they used only underlying patient information or static patient disease information. In this study, we propose a multilayer perceptron (MP)-based IDH prediction model using heart rate (HR) information corresponding to time-series information and static data of patients. This study aimed to validate whether HR differences and HR slope information affect real-time IDH prediction in patients undergoing hemodialysis. Clinical data were collected from 80 hemodialysis patients from 9 September to 17 October 2020, in the artificial kidney room at Yeungnam University Medical Center (YUMC), Daegu, South Korea. The patients typically underwent hemodialysis 12 times during this period, 1 to 2 h per session. Therefore, the HR difference and HR slope information within up to 1 h before IDH occurrence were used as time-series input data for the MP model. Among the MP models using the number and data length of different hidden layers, the model using 60 min of data before the occurrence of two layers and IDH showed maximum performance, with an accuracy of 81.5%, a true positive rate of 73.8%, and positive predictive value of 87.3%. This study aimed to predict IDH in real-time by continuously supplying HR information to MP models along with static data such as age, diabetes, hypertension, and ultrafiltration. The current MP model was implemented using relatively limited parameters; however, its performance may be further improved by adding additional parameters in the future, further enabling real-time IDH prediction to play a supporting role for medical staff.

Keywords: heart-rate; hemodialysis; intradialytic hypotension; multilayer perceptron; real-time.

Figure 1

(a) normal and (b) inadequate compensatory responses to maintain BP during dialysis ([14]). (In Figure 1a, + denotes the response sensitivity.)

Figure 2

Pathophysiology of IDH ([33]).

Figure 3

Proposed MP-IDH net with static and dynamic data inputs.

Figure 4

Changes in slope of HR per minute before IDH. ((a–h,I,j) show the decreasing and increasing trend of the mean HR slope respectively).

Figure 5

Relationship between BP and baroreceptor reflex ([39]).

Figure 6

Distribution of HR differences for (a) IDH and (b) normal (non-IDH) patients 1 h before IDH onset.

Figure 7

Changes in HR slopes of 1 h, 45 min, 30 min, and 15 min data for IDH and normal (Non-IDH) patients before IDH onset. HR slopes before IDH onset for (a) IDH and (b) normal (Non-IDH) patients. (bold color line: average HR slope, black line: x-axis).

Figure 8

Changes in HR slope for IDH and normal (non-IDH) patients according to 30-min data by patient baseline information before the onset of IDH. (The dashed lines represent the lines separating the patient baseline information).

Figure 9

Confusion matrices of Deep-IDH models using different hidden layers and data lengths: (a) 1-layer and 60-min data before IDH occurrence (69.2%, 30.8%); (b) 1-layer and 45-min data before IDH onset (73.2%, 26.8%); (c) 1-layer and 30-min data before IDH onset (64.5%, 35.5%); (d) 2-layer and 60-min data before IDH onset (81.5%, 18.5%); (e) 2-layer and 45-min data before IDH onset (70.2%, 29.8%); and (f) 2-layer and 30-min data before IDH onset (59.6%, 40.4%).

Figure 9

Confusion matrices of Deep-IDH models using different hidden layers and data lengths: (a) 1-layer and 60-min data before IDH occurrence (69.2%, 30.8%); (b) 1-layer and 45-min data before IDH onset (73.2%, 26.8%); (c) 1-layer and 30-min data before IDH onset (64.5%, 35.5%); (d) 2-layer and 60-min data before IDH onset (81.5%, 18.5%); (e) 2-layer and 45-min data before IDH onset (70.2%, 29.8%); and (f) 2-layer and 30-min data before IDH onset (59.6%, 40.4%).

Figure 10

ROC of Deep-IDH models using different hidden layers and data lengths: (a) 1-layer and 60-min data before IDH onset; (b) 1-layer and 45-min data before IDH onset; (c) 1-layer and 30-min data before IDH onset; (d) 2-layer and 60-min data before IDH onset; (e) 2-layer and 45-min data before IDH onset; and (f) 2-layer and 30-min data before IDH onset.

Figure 11

HRV results at the time of hypotension in hemodialysis patients.

Figure 12

Changes in HR per minute at the onset of IDH in hemodialysis patients. (The square box area indicates the IDH or hypotension period).