運用類神經網路建構腎結石的預測模型

Using Artificial Neural Networks to Build Prediction Models of Kidney Stones

陳智堂(Chih-Tang Chen)

指導教授 ： 徐建業

臺北醫學大學/醫學科技學院/醫學資訊研究所/碩士(2012年)

https://doi.org/10.6831/TMU.2012.00082

摘要

到目前腎結石是否容易由某些疾病引起已有諸多推測，我們找出五種可能導致腎結石的疾病，我們用大量資料樣本並以加以分析使用了T test同時也使用類神經網路及二元Logistic regression建置預測模型看是否能預測腎結石發生。
本研究以國民健康保險局的百萬歸人檔中2000~2003的病患作為樣本，過濾在2000~2003間有罹患腎結石的病患，過濾剩下的病患則為我們的樣本並且統計樣本中的病患在2000~2003有無罹患糖尿病、副甲狀腺機能亢進、痛風關節炎、高血脂及高血壓，接著我們再從國民健康保險局的百萬歸人檔中2004~2008的資料中去追蹤樣本中的病患是否再這幾年間有得到腎結石，接著對資料做敘述性分析統計，最後再以STATISTCA 8的Automated Neural Network工具對樣本訓練及建構類神經網路模型及預測，並且嘗試使用較少的疾病因子去訓練預測模型，我們也以二元Logistic regression對樣本做預測的測試。
最後本研究發現，本研究以所有變數去訓練出的類神經模型可以得到最佳ROC面積，而且本研究的樣本相當大使得由本研究所產生的統計數據具有很高的參考價值。

關鍵字

腎結石、類神經網路

並列摘要

So far, there are lots of speculations of certain diseases could cause the formation of kidney stones, we find out five disease which could contribute to the Formation of kidney stones and we use artificial neural networks to build up prediction model and compare with binary logistic regression, at the same time, we also try to use less factors of diseases to aim the prediction.
This research use National Health Insurance Research Database(NHIRD) as data source, we selected patients who have not been diagnosed any kidney stone between 2000 and 2003 and we counted whether they have been diagnosed diabetes mellitus、Gouty Arthropathy、hyperlipidemia and Hypertension , then we tracked them from 2004 to 2008 to see whether they were diagnosed kidney stones as result. After this procedure we use automated neural network tool of STATISTICA 8 to build up a prediction model for the data and we attempt to use less factor to build the prediction model which can attain the same or greater prediction ability. We also use binary logistic regression to verified the artificial neural network model that trained by us.
Finally, the artificial neural network model with all variables trained by us can get the best performance. Plus, the samples we use in this research is quite large so it makes the statistical number of this research is trustful and has a high value for reference.

並列關鍵字

kidney stones、artificial neural network

參考文獻

27. Sanchez-Martin FM, Millan Rodriguez F, Esquena Fernandez S, et al. (2007). Incidence and prevalence of published studies about urolithiasis in Spain: a review [article in Spanish] Actas Urol Esp. 2007,31, 511–520.

2. Aloua Akoudad, Moyses Szklo, Mara A. McAdams, Tibor Fulop, Cheryl A.M. Anderson, Josef Coresh, Anna Kottgen.(2010).Correlates of Kidney stone disease differ by race in a multi-ethnic middle-aged population: The ARIC study.Preventive Medicine, 51, 416-420.

3. Amato M, Lusini ML, Nelli F. (2004). Epidemiology of nephrolithiasis today. Urol Int. 2004, 72(suppl 1), 1–5.

5. Chi Pang Wen MD ,Ting Yuan David Cheng MS, Min Kuang Tsai MS, Yen Chen Chang MS, Hui Ting Chan MS, Shan Pou Tsai PhD,et al. (2008). All-cause mortality attributable to chronic kidney disease: a prospective cohort study based on 462 293 adults in Taiwan.The Lacent, Volume 371, Issue 9631, 28 June-4 July 2008, 2173-2182.

7. Chen C-L, Kamatani N, Nishioka K, Mikanagi K.(1989). Clinical aspects of gouty patients in Taiwan. Adv Exp Med Biol, 253A, 189–95.