Prediction of hemodialysis treatment results with neural networks and two-compartment model

• Autorzy: Korohoda P. , Grabska-Chrząstowska J.

Warianty tytułu

PL

Przewidywanie wyników leczenia hemodializą z wykorzystaniem sieci neuronowych i modelu dwuprzedziałowego

• Języki publikacji: EN

Abstrakty

EN

In the paper an experiment is described, designed and conducted to verify hypothesis that neural networks of relatively small size can provide relevantly accurate approximation of unknown analytical solution of the two-compartment hemodialysis model with variable volume of the extracellular compartment. The experiment was based on a thousand pseudorandomly generated sessions. The concentration values at the end of treatment and the equilibrated Kt/V index were approximated with absolute relative error less or equal 2% for around 90% values, with only 8 or 12 hidden neurons. Such results have been considered to confirm the main hypothesis.

PL

W artykule opisano eksperyment, zaprojektowany i zrealizowany w celu zweryfikowania hipotezy, że relatywnie nieduża sieć neuronowa może zapewnić odpowiednio dokładną aproksymację nieznanego rozwiązania analitycznego dla modelu dwuprzedziałowego ze zmienną objętością przedziału zewnątrzkomórkowego. Eksperyment bazował na tysiącu pseudolosowo wygenerowanych sesji. Wartości stężeń w chwili zakończenia zabiegu oraz indeks Kt/V dla uśrednionego stężenia były oszacowane z dokładnością modułu błędu względnego mniejszą lub równą 2% dla około 90% wartości, przy zaledwie 8 lub 12 neuronach w warstwie ukrytej. Wynik ten uznano za potwierdzający prawdziwość postawionej hipotezy.

Słowa kluczowe

EN

multilayer perceptron network; hemodialysis modelling; two-pool model

PL

wielowarstwowa sieć perceptronowa; modelowanie hemodializy; model dwuprzedziałowy

• Wydawca: Wydawnictwa AGH
• Czasopismo: Automatyka / Akademia Górniczo-Hutnicza im. Stanisława Staszica w Krakowie
• Rocznik: 2011
• Tom: T. 15, z. 3
• Strony: 583--593
• Opis fizyczny: Bibliogr. 29 poz., rys., wykr., tab.

Twórcy

autor

Korohoda P.

• AGH University of Science and Technology, Faculty of Electrical Engineering Automatics, Computer Science and Electronics, Department of Electronics, al. A. Mickiewicza 30, 30-059 Kraków, Poland

autor

Grabska-Chrząstowska J.

• AGH University of Science and Technology, Faculty of Electrical Engineering Automatics, Computer Science and Electronics, Department of Automatics, al. A. Mickiewicza 30, 30-059 Kraków, Poland

Bibliografia

• [1] Akl A.I., Sobh M.A., Enab Y.M., Tattersall J., Artificial intelligence: A new approach for prescription and monitoring of hemodialysis therapy. American Journal of Kidney Diseases, 38 (6), 001, 1277-1283.
• [2] Arnold V., Ordinary differential eąuations. Springer, 1992.
• [3] Azar A., Wahba D., Association Between Neural Network And System Dynamics To Predict Dialysis Dose During Hemodialysis. The 2008 International Conference of the System Dynamics ociety, 2008, Greece.
• [4] Daugirdas J.T., Tattersall J., Effect of treatment spacing and freąuency on three measures of quivalent clearance, including standard Kt/V. Nephrol Dial Transplant, 25, 2010, 558-561.
• [5] Depner TA., Prescribing Hemodialysis: A Guide to Urea Modeling. Kluwer Academic Publishers, 6th print, 1997.
• [6] Fausett L., Fundamentals of Neural Networks. Prentice-Hall, New Jersey 1994.
• [7] Fernandez, E.A., Valtuille, R., Willshaw, P , Perazzo, C.A.: Dialysate-side urea kinetics. Neural etwork predicts dialysis dose during dialysis. Medical and Biological Engineering and Computing, vol. 41, Issue 4, July 2003, 392-396.
• [8] Fernandez E.A., Valtuille R., Presedo J.R., Willshaw P., Comparison of standard and artificial neural network estimators of hemodialysis adeąuacy. Artificial Organs, vol. 29, Issue 2, February 2005, 159-165.
• [9] Gabutti L., Vadilonga D., Mombelli G., Burnier M., Marone C, Artificial neural networks improve the prediction of Kt/V, follow-up dietary protein and hypotension risk in haemodialysis patients. Nephrology Dialysis and Transplantation, vol. 19, Issue 5, 2004, 1204-1211.
• [10] Goldfarb-Rumyantzev A., Schwenk M.H., Liu S., Charytan C, Spinowitz B.S., Prediction of single-pool Kt/V based on clinical and hemodialysis variables using multilinear regression, tree-based modeling, and artificial neural networks. Artificial Organs vol. 27, Issue 6, 1 June 2003, 544-554.
• [11] Gotch F.A., Erolution ofthe Single-Pool Urea Kinetic Model. Seminars in Dialysis, 14(4), 2001, 252-256.
• [12] Grandi F., Avanzolini G., Cappello A., Analytic solution ofthe variable-volume double-pool urea kinetic model applied to parameter estimation in hemodialyisis. Computers in Biology and Medicine, 25(6), 1995, 505-518.
• [13] Korohoda P., Hemodialysis modeling based on measurement data — optimization procedurę based on two-compartment model. Automatyka (półrocznik AGH), 11(3), 2007, 179-184 (in Polish).
• [14] Korohoda P., SchneditzD., Pietrzyk J.A., Hemodialysis modeling - the most often used kidney replacement therapy. A chapter in monography: Bioengineering - basics, vol. 2, edited by R. Tadeusiewicz and P. Augustyniak, Kraków, Wydawnictwa AGH 2009 (in Polish).
• [15] Korohoda P., Simplified flow-based hemodialysis model - comparison with classic two-compartment model. Automatyka (półrocznik AGH), 13, 3, 2009, 1129-1140 (in Polish).
• [16] Korohoda P., Pietrzyk J.A., Sułowicz W., Weekly based hemodialysis dose indicator KT/V: new possibility to assess efficiency of dialysis in non-conventional treatment schemes. Nefrologia i Dializoterapia Polska, 13(3), 2009, 138-142 (in Polish).
• [17] Korohoda P., Grabska-Chrząstowska J., Application of neural networks in urea kinetic modeling. Automatyka (półrocznik AGH), 14(3/2), 2010, 785-793 (in Polish).
• [18] Korohoda P., Schneditz D., Pietrzyk J.A., Quantifying the discontinuity of haemodialysis dose with time-averaged concentration (TAC) and time-averaged deviation (TAD). Nephrology Dialysis Transplantation, vol. 25 No. 3, 2010, 1011-1012.
• [19] Pietrzyk J.A., Kinetic modeling ofurea. DWN DReAM, Kraków, 1992 (in Polish).
• [20] Press W.H., Teukolsky S.A., Vetterling W.T., Flannery B.P., Numerical Recipes in C. 2nd ed.,Cambridge Univ. Press, 1992.
• [21] Schneditz D., Daugirdas J.T., Formal analytical solution to a regional blood flow and diffusion based urea kinetic model. ASAIO Journal, 40, 1994, M667-673.
• [22] Smye S.W, Will E.J., A mathematical analysis ofa two-compartment model ofurea kinetics. Phys. Med. Biol., 40, 1995, 2005-2014.
• [23] Szmajda R., Szczepaniak PS., Modeling hemodialysis: regression versus neural model. Journal of Medical Informatics and Technologies, vol. 7, 2004.
• [24] Tadeusiewicz R., Neural networks. Akademicka Oficyna Wydawnicza RM, Warszawa 1993 (inPolish).
• [25] Tadeusiewicz R., Using Neural Models for Evaluation of Biological Activity of Selected Chemical Compounds. Chapter (6) in book: Smolinski T.G, Milanova M.G. and Hassanien A.-E. (Eds.) Applications of Computational Intelligence in Biology, Current Trends and Open Problems, Studies in Computational Intelligence, vol. 122, Springer-Verlag, Berlin - Heidelberg - New York, 2008, 135-159.
• [26] Tadeusiewicz R., Using Neural Networks for Simplified Discovery ofSome Psychological Phenomena. Chapter in book: Rutkowski L. (et al., eds.): Artificial Intelligence and Soft Computing, Springer-Verlag, Berlin - Heidelberg - New York, 2010, 104-123.
• [27] Zurada J.M., Artificial Neural Systems. PWS Publishing Company, 1992.
• [28] Matlab www page: http://www.mathworks.com/.
• [29] Statistica www page: http://www.statsoft.pl/.