Self-Adaptive Artificial Neural Networks Applied to Brain Death Level and Antibiotics Treatment of Gastric Lymphoma Cancer Prognostication

Self-Adaptive Artificial Neural Networks Applied to Brain Death Level and Antibiotics Treatment of Gastric Lymphoma Cancer Prognostication

莫哈馬(Muammar Sadrawi)

指導教授 ： 謝建興

元智大學/工程學院/機械工程學系/碩士(2013年)

https://doi.org/10.6838/YZU.2013.00031

摘要

These studies evaluated the applications of Artificial Neural Networks (ANNs) for being used in brain death patient level and gastric lymphoma cancer antibiotics. Back-propagation neural network (BPNN) has been applied to make the prediction beside the fuzzy model for brain death patient level. The model is influenced by 12 inputs. The results showed the self-adaptive ensembled neural networks (SeA-EANN) gave the best model followed by manually tuned fuzzy modeling, ensembled neuro fuzzy inference system (EANFIS) and ensembled neural networks (EANN), which produced testing MSE 0.00845, 0.019, 0.021 and 0.026, respectively. Another case, using single SeA-ANN, as the model for distinguishing resistive and sensitive patient to antibiotics also has been developed. We used multiple sensitivity analysis in choosing the most important genes, from 654 to 50. For the result, we got 90% accuracy and several genes can be used for further clinical validation.

關鍵字

Brain death level ； Fuzzy modeling ； Artificial neural networks ； Gastric lymphoma cancer antibiotics ； Sensitivity analysis

並列摘要

These studies evaluated the applications of Artificial Neural Networks (ANNs) for being used in brain death patient level and gastric lymphoma cancer antibiotics. Back-propagation neural network (BPNN) has been applied to make the prediction beside the fuzzy model for brain death patient level. The model is influenced by 12 inputs. The results showed the self-adaptive ensembled neural networks (SeA-EANN) gave the best model followed by manually tuned fuzzy modeling, ensembled neuro fuzzy inference system (EANFIS) and ensembled neural networks (EANN), which produced testing MSE 0.00845, 0.019, 0.021 and 0.026, respectively. Another case, using single SeA-ANN, as the model for distinguishing resistive and sensitive patient to antibiotics also has been developed. We used multiple sensitivity analysis in choosing the most important genes, from 654 to 50. For the result, we got 90% accuracy and several genes can be used for further clinical validation.

並列關鍵字

Brain death level ； Fuzzy modeling ； Artificial neural networks ； Gastric lymphoma cancer antibiotics ； Sensitivity analysis

參考文獻

1.Bishop, C. M., “Neural Networks for Pattern Recognition”. Clarendon Press, Oxford, 1995.

3.Rosendale, J. D., Kauffman, H. M., McBride,M.A., Chabalewski,F.L., Zaroff,J.G., Garrity,E.R., Delmonico,F.L., and Rosengard,B.R., “Aggressive Pharmacological Donor Management Result in More Transplanted Organs,” Transplantation, 75, 482, 2003.

4.Health Resource and Service Administration (HRSA), http://optn.transplant.hrsa.gov/latestData/rptData.asp, accessed Sept 6th, 2012.

5.Mizraji,R., Perezz-Protto,S., Etchegaray, A., Castro, A., Lander, M., Buccino, E., Severo, L., and Alvarez, I., “Brain Death Epidemiology in Uruguay and Utilization of the Glasgow Coma Score in Acute Brain Injured Patients as a Predictor of Brain Death,”Transpl. Proc., 41, 3489, 2009.

7.Machado, C., “Diagnosis of Brain Death,”Neurology International, Vol. 2:e2, 2010.