According to the 2018 annual report of US Department of Kidney Data System (USRDS), Taiwan's dialysis rate and prevalence rate are the highest in the world due to population aging, diabetes and progresses in cardiovascular care. With the rise of artificial intelligence deep learning in recent years, various analytical software resources have gradually become easier to obtain. At the same time, wearable cyber physical sensors are becoming more and more popular. Measurements on vital signs such as heartbeat, electrocardiogram, and blood oxygenation blood pressure values are ubiquitous. We propose an integrated system that combines dialysis big data deep learning analysis with cross platform physiological sensing. We specifically tackle the early warning of dialysis discomfort such as hypotension, hypertension, cramps, etc., this requires a large amount of data collection, related training, data sources including dialysis treatment process and home physiological data. Although the Dialysis machine is able to produce huge amount of IoT data, the usable data for early warning system training is not as huge due to the limited physician labors devoted for labeling questionable samples. This generally leads to low accuracy for regular CNN training methods. We enhance the AI training performance via a transfer learning technique. The AI training accuracy reaches the value of 99% with the help of transfer learning, while that of an original CNN process on the HD data bears a low 60% accuracy. Given the high prediction accuracy of our AI engine, we are able to integrate the real time measurements from Dialysis machine with wearable devices such as ECG sensors and wrist health watches, and make precision prediction of incoming discomfort during the HD treatments. The ECG signal of the same group patients are also analyzed with the same technique. The same accuracy enhancement are also observed.
[1]
José Machado,et al.
Understanding Stroke in Dialysis and Chronic Kidney Disease
,
2017,
EUSPN/ICTH.
[2]
Min Zhou,et al.
ECG Classification Using Wavelet Packet Entropy and Random Forests
,
2016,
Entropy.
[3]
Yuh-Feng Lin,et al.
Applying an Artificial Neural Network to Predict Total Body Water in Hemodialysis Patients
,
2005,
American Journal of Nephrology.
[4]
Wei-Hsin Huang,et al.
Prediction of target range of intact parathyroid hormone in hemodialysis patients with artificial neural network
,
2006,
Comput. Methods Programs Biomed..
[5]
M. Brier,et al.
Artificial intelligence for optimal anemia management in end-stage renal disease.
,
2016,
Kidney international.
[6]
U. Rajendra Acharya,et al.
ECG beat classification using PCA, LDA, ICA and Discrete Wavelet Transform
,
2013,
Biomed. Signal Process. Control..
[7]
Jürgen Schmidhuber,et al.
A reinforcement learning approach for individualizing erythropoietin dosages in hemodialysis patients
,
2009,
Expert Syst. Appl..
[8]
U. Rajendra Acharya,et al.
A deep convolutional neural network model to classify heartbeats
,
2017,
Comput. Biol. Medicine.
[9]
Ahmad Taher Azar,et al.
Modeling and Control of Dialysis Systems
,
2012
.