透析中低血壓 (Intradialytic Hypotension, IDH) 是血液透析療程中常見的併發症，IDH 的發生率介於 5% 至 30% 之間；此病症可誘發病人產生頭暈、噁心和嘔吐等不良反應，嚴重情況時更會導致昏厥、休克甚至死亡。IDH 之發生為透析患者的存活重要風險因子，相較沒有發生 IDH 的患者，好發 IDH 的患者其死亡風險增加了 1.56 倍，對臨床實務來說，準確預測 IDH 的風險以及對預測結果的解釋性至關重要。目前臨床上對於 IDH 的預測主要依賴過往照護經驗與對患者當下病發之表徵進行判斷；因人工智慧的蓬勃發展，近年來也有研究探討使用機器學習與深度學習來進行 IDH 的預測，但在準確性和可解釋性上仍有改進的空間。為此本研究提出動態調整權重的集成學習方法用於預測 IDH 風險，研究資料來自臺南市立安南醫院血液透析室，共收集自502 名患者的 121,844 次血液透析療程。利用多種機器學習分類器，包括Random Forest、Extremely Randomized Tree、Logistic Regression、Light Gradient Boosting Machine、eXtreme Gradient Boosting 和 Adaptive Boosting；再經過超參數搜索和權重計算後，將結果輸入到投票分類器進行預測，權重根據各分類器於驗證集中的效能來動態調整從而提高預測效能。結果顯示以 Area Under Precision-Recall Curve 為權重的投票分類器超越單一機器學習分類器之成效，該模型在 Area Under Receiver Operating Characteristic Curve達 93.54% ± 0.02，並取得 88.60% ± 0.08 Sensitivity，與 83.59% ± 0.06 Specificity 的成效。結果進一步證實：SHapley Additive exPlanations、特徵重要性，以及統計分析與臨床研究所揭示的 IDH 風險因子均具有一致性，此發現彰顯本研究之模型在訓練和預測過程中，能有效地從患者特徵資料中學習到 IDH 發生之形式，顯示本研究之方法不僅在臨床上具備應用性和解釋性，為醫護人員提供對模型預測過程更深入理解，提供基於資料決策的透析療程設計，進而優化治療策略以持續降低 IDH 的發生率、改善患者之透析品質。

Intradialytic Hypotension (IDH) frequently occurs during hemodialysis, leading to symptoms like dizziness and, in extreme cases, even death. With an incidence rate between 5% and 30%, IDH significantly increases the mortality risk by 1.56 times. Although current IDH predictions use past clinical experiences and patient symptoms, the emergence of artificial intelligence offers novel predictive methodologies. This research employed a dynamically weighted ensemble learning approach to predict IDH risk using data from 121,844 hemodialysis sessions of 502 patients from Tainan City An-nan Hospital. Several machine learning classifiers were utilized, undergoing hyperparameter tuning and weight calculation. The weights, based on each classifier's validation set performance, were then adjusted to enhance prediction efficacy. Results revealed that our model, utilizing a voting classifier weighted by the Area Under Precision-Recall Curve, surpassed other individual classifiers, achieving 93.54% in the Area Under the Receiver Operating Characteristic Curve with 88.60% sensitivity and 83.59% specificity. The consistency observed between the model’s feature importance and clinically recognized IDH risk factors highlights the model's clinical relevance, aiding healthcare professionals in understanding and designing data-driven hemodialysis strategies to optimize patient outcomes.

[1] V. Jha et al., “Chronic kidney disease: Global dimension and perspectives,” Lancet, vol. 382, no. 9888, pp. 260–272, 2013.
[2] P. Teresa K. Chen, MD, MHS, Daphne H. Knicely, MD, Morgan E. Grams, MD, “Chronic Kidney Disease Diagnosis and Management,” HHS Public Access, 2019.
[3] M. E. Grams et al., “Race, APOL1 risk, and EGFR decline in the general population,” J. Am. Soc. Nephrol., vol. 27, no. 9, pp. 2842–2850, 2016.
[4] T. Y. Lin, J. S. Liu, and S. C. Hung, “Obesity and risk of end-stage renal disease in patients with chronic kidney disease: A cohort study,” Am. J. Clin. Nutr., vol. 108, no. 5, pp. 1145–1153, 2018.
[5] W. W. C. de B. I. H. R. P. Jadoul Michel, “KDIGO Clinical Practice Guideline on Diabetes Mangement in Chronic Kidney Disease,” Kdigo Clin. Pract. Guidel. Diabetes Mangement Chronic Kidney Dis., no. 12, pp. 1–152, 2022.
[6] Hashmi MF, Benjamin O, Lappin SL. End-Stage Renal Disease. NCBI, 2023.
[7] E. Zitt et al., “Long-term risk for end-stage kidney disease and death in a large population-based cohort,” Sci. Rep., vol. 8, no. 1, pp. 1–8, 2018.
[8] N. H. R. I. & T. S. of Nephrology, 2021 Annual Report on Kidney Disease in Taiwan. 2021.
[9] R. Sinnakirouchenan and J. L. Holley, “Peritoneal dialysis versus hemodialysis: Risks, benefits, and access issues,” Adv. Chronic Kidney Dis., vol. 18, no. 6, pp. 428–432, 2011.
[10] T. S. Lai, C. C. Hsu, M. H. Lin, V. C. Wu, and Y. M. Chen, “Trends in the incidence and prevalence of end-stage kidney disease requiring dialysis in Taiwan: 2010–2018,” J. Formos. Med. Assoc., vol. 121, pp. S5–S11, 2022.
[11] M. A. Shafiee, P. Chamanian, P. Shaker, Y. Shahideh, and B. Broumand, “The impact of hemodialysis frequency and duration on blood pressure management and quality of life in end-stage renal disease patients,” Healthc., vol. 5, no. 3, 2017.
[12] C. Zoccali, F. A. Benedetto, G. Tripepi, and F. Mallamaci, “Cardiac consequences of hypertension in hemodialysis patients,” Semin. Dial., vol. 17, no. 4, pp. 299–303, 2004.
[13] J. Kuipers et al., “The Prevalence of Intradialytic Hypotension in Patients on Conventional Hemodialysis: A Systematic Review with Meta-Analysis,” Am. J. Nephrol., vol. 49, no. 6, pp. 497–506, 2019.
[14] B. Sars, F. M. Van Der Sande, and J. P. Kooman, “Intradialytic Hypotension: Mechanisms and Outcome,” Blood Purif., vol. 49, no. 1–2, pp. 158–167, 2019.
[15] T. Shoji, Y. Tsubakihara, M. Fujii, and E. Imai, “Hemodialysis-associated hypotension as an independent risk factor for two-year mortality in hemodialysis patients,” Kidney Int., vol. 66, no. 3, pp. 1212–1220, 2004.
[16] J. J. Sands et al., “Intradialytic hypotension: Frequency, sources of variation and correlation with clinical outcome,” Hemodial. Int., vol. 18, no. 2, pp. 415–422, 2014.
[17] R. F. Reilly, “Attending rounds: A patient with intradialytic hypotension,” Clin. J. Am. Soc. Nephrol., vol. 9, no. 4, pp. 798–803, 2014.
[18] P. N. Van Buren, “Redefining the Burden of Intradialytic Hypotension in the Modern Era of Hemodialysis,” Am. J. Nephrol., vol. 49, no. 6, pp. 494–496, 2019.
[19] O. C. Okoye, H. E. Slater, and N. Rajora, “Prevalence and risk factors of intra-dialytic hypotension: A 5 year retrospective report from a single Nigerian center,” Pan Afr. Med. J., vol. 28, pp. 1–7, 2017.
[20] A. Gul, D. Miskulin, A. Harford, and P. Zager, “Intradialytic hypotension,” Current Opinion in Nephrology and Hypertension, vol. 25, no. 6. Lippincott Williams and Wilkins, pp. 545–550, Nov. 01, 2016.
[21] H. Hirakata et al., “Japanese Society for Dialysis Therapy Guidelines for Management of Cardiovascular Diseases in Patients on Chronic Hemodialysis,” Ther. Apher. Dial., vol. 16, no. 5, pp. 387–435, 2012.
[22] M. M. Assimon and J. E. Flythe, “Definitions of intradialytic hypotension,” Seminars in Dialysis, vol. 30, no. 6, pp. 464–472, 2017.
[23] L. J. Chesterton, N. M. Selby, J. O. Burton, and C. W. McIntyre, “Cool dialysate reduces asymptomatic intradialytic hypotension and increases baroreflex variability,” Hemodial. Int., vol. 13, no. 2, pp. 189–196, 2009.
[24] J. E. Flythe, H. Xue, K. E. Lynch, G. C. Curhan, and S. M. Brunelli, “Association of mortality risk with various definitions of intradialytic hypotension,” J. Am. Soc. Nephrol., vol. 26, no. 3, pp. 724–734, 2015.
[25] B. V. Stef Á Nsson et al., “Intradialytic hypotension and risk of cardiovascular disease,” Clin. J. Am. Soc. Nephrol., vol. 9, no. 12, pp. 2124–2132, 2014.
[26] R. Matsuura et al., “Intradialytic hypotension is an important risk factor for critical limb ischemia in patients on hemodialysis,” BMC Nephrol., vol. 20, no. 1, pp. 1–8, 2019.
[27] M. Kanbay et al., “An update review of intradialytic hypotension: Concept, risk factors, clinical implications and management,” Clin. Kidney J., vol. 13, no. 6, pp. 981–993, 2020.
[28] R. Raina et al., “Pediatric intradialytic hypotension: recommendations from the Pediatric Continuous Renal Replacement Therapy (PCRRT) Workgroup,” Pediatr. Nephrol., vol. 34, no. 5, pp. 925–941, 2019.
[29] W. Chen, F. Wang, Y. Zhao, L. Zhang, Z. Chen, and M. Dai, “Efficacy and safety of furosemide for prevention of intradialytic hypotension in haemodialysis patients: Protocol for a multicentre randomised controlled trial,” BMJ Open, vol. 11, no. 7, 2021.
[30] A. M. L. S. M. Christopher, “A Brief Review of Intradialytic Hypotension with a Focus on Survival,” Physiol. Behav., vol. 176, no. 1, pp. 100–106, 2016.
[31] A. Brnabic and L. M. Hess, “Systematic literature review of machine learning methods used in the analysis of real-world data for patient-provider decision making,” BMC Med. Inform. Decis. Mak., vol. 21, no. 1, 2021.
[32] M. K. Siddiqui, R. Morales-Menendez, X. Huang, and N. Hussain, “A review of epileptic seizure detection using machine learning classifiers,” Brain Informatics, vol. 7, no. 1, pp. 1–18, 2020.
[33] L. Tian et al., “Radiomics-based machine-learning method for prediction of distant metastasis from soft-tissue sarcomas,” Clin. Radiol., vol. 76, no. 2, pp. 158.e19-158.e25, 2021.
[34] I. H. Sarker, “Machine Learning: Algorithms, Real-World Applications and Research Directions,” SN Comput. Sci., vol. 2, no. 3, pp. 1–21, 2021.
[35] T. G. Dietterich, “Machine-Learning Research”, AIMag, vol. 18, no. 4, p. 97, Dec. 1997.
[36] L. K. Hansen and P. Salamon, “Neural network ensembles,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 12, October, pp. 993–1001, 1990.
[37] Dietterich, T.G. Author of Part, "Ensemble Methods in Machine Learning," in Multiple Classifier Systems: MCS 2000. Lecture Notes in Computer Science, Vol. 1857, Springer, Ed. Berlin, Heidelberg: Springer, 2000, pp. 1–15.
[38] X. Dong, Z. Yu, W. Cao, Y. Shi, and Q. Ma, “A survey on ensemble learning,” Front. Comput. Sci., vol. 14, no. 2, pp. 241–258, 2020.
[39] W. H. Beluch, T. Genewein, A. Nurnberger and J. M. Kohler, "The Power of Ensembles for Active Learning in Image Classification," 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018, pp. 9368-9377
[40] T. Zhou, H. Lu, Z. Yang, S. Qiu, B. Huo, and Y. Dong, “The ensemble deep learning model for novel COVID-19 on CT images,” Appl. Soft Comput., vol. 98, p. 106885, 2021.
[41] P. Singh, “Comparative study of individual and ensemble methods of classification for credit scoring,” Proc. Int. Conf. Inven. Comput. Informatics, ICICI 2017, Icici, pp. 968–972, 2018.
[42] S. Li, X. Lu, S. Sakai, M. Mimura and T. Kawahara, "Semi-supervised ensemble DNN acoustic model training," 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, USA, 2017, pp. 5270-5274,
[43] F. Huang, G. Xie, and R. Xiao, “Research on ensemble learning,” 2009 Int. Conf. Artif. Intell. Comput. Intell. AICI 2009, vol. 3, pp. 249–252, 2009.
[44] M. A. Ganaie, M. Hu, A. K. Malik, M. Tanveer, and P. N. Suganthan, “Ensemble deep learning: A review,” Engineering Applications of Artificial Intelligence, vol. 115. 2022.
[45] J. T. Kwok, Z. H. Zhou, and L. Xu, "Machine Learning," in Springer Handbook of Computational Intelligence, J. Kacprzyk and W. Pedrycz, Eds. Berlin, Heidelberg: Springer, 2015, pp. 495-522.
[46] LEO BREIMAN, “Random Forests,” Mach. Learn., vol. 45, no. 5–32, pp. 1–122, 2001.
[47] L. Li, Q. Hu, X. Wu, and D. Yu, “Exploration of classification confidence in ensemble learning,” Pattern Recognit., vol. 47, no. 9, pp. 3120–3131, 2014.
[48] P. Sarder, B. Ginley, and J. E. Tomaszewski, “Automated renal histopathology: digital extraction and quantification of renal pathology,” Med. Imaging 2016 Digit. Pathol., vol. 9791, March 2016, p. 97910F, 2016.
[49] S. Sheehan, S. Mawe, R. E. Cianciolo, R. Korstanje, and J. M. Mahoney, “Detection and Classification of Novel Renal Histologic Phenotypes Using Deep Neural Networks,” Am. J. Pathol., vol. 189, no. 9, pp. 1786–1796, 2019.
[50] N. Tangri et al., “A predictive model for progression of chronic kidney disease to kidney failure,” Jama, vol. 305, no. 15, pp. 1553–1559, 2011.
[51] M. Shahabi, V. R. Nafisi, and F. Pak, “Prediction of intradialytic hypotension using PPG signal features,” 2015 22nd Iran. Conf. Biomed. Eng. ICBME 2015, no. November, pp. 399–404, 2016.
[52] C. J. Lin et al., “Intelligent system to predict intradialytic hypotension in chronic hemodialysis,” J. Formos. Med. Assoc., vol. 117, no. 10, pp. 888–893, 2018.
[53] H. Lee et al., “Deep learning model for real-time prediction of intradialytic hypotension,” Clin. J. Am. Soc. Nephrol., vol. 16, no. 3, pp. 396–406, 2021.
[54] H. W. Kim et al., “Deep Learning Model for Predicting Intradialytic Hypotension Without Privacy Infringement: A Retrospective Two-Center Study,” Front. Med., vol. 9, July 2022.
[55] J. Kuipers et al., “Association between quality of life and various aspects of intradialytic hypotension including patient-reported intradialytic symptom score,” BMC Nephrol., vol. 20, no. 1, pp. 1–8, 2019.
[56] J. -Y. Yang et al., "Differencing Time Series as an Important Feature Extraction for Intradialytic Hypotension Prediction using Machine Learning," 2021 IEEE 3rd Eurasia Conference on Biomedical Engineering, Healthcare and Sustainability (ECBIOS), Tainan, Taiwan, 2021, pp. 19-20
[57] H. Zhang et al., “Real-Time Prediction of Intradialytic Hypotension using Machine Learning and Cloud Computing Infrastructure,” Nephrology Dialysis Transplantation, 2023.
[58] D. F. Keane, J. G. Raimann, H. Zhang, J. Willetts, S. Thijssen, and P. Kotanko, “The time of onset of intradialytic hypotension during a hemodialysis session associates with clinical parameters and mortality,” Kidney Int., vol. 99, no. 6, pp. 1408–1417, 2021.
[59] C. Y. J. Peng, K. L. Lee, and G. M. Ingersoll, “An introduction to logistic regression analysis and reporting,” J. Educ. Res., vol. 96, no. 1, pp. 3–14, 2002.
[60] Y. Freund, R. E. Schapire, P. Avenue, and F. Park, “A Short Introduction to Boosting,” J. Japanese Soc. Artif. Intell., vol. 14, no. 5, pp. 771–780, 1999.
[61] L. Breiman, “Random Forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, Oct. 2001.
[62] P. Geurts, D. Ernst, and L. Wehenkel, “Extremely randomized trees,” Mach. Learn., vol. 63, no. 1, pp. 3–42, 2006.
[63] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” Proc. ACM SIGKDD Int. Conf. Knowl. Discov. Data Min., vol. 13-17-Augu, pp. 785–794, 2016.
[64] G. Ke et al., “LightGBM: A highly efficient gradient boosting decision tree,” Adv. Neural Inf. Process. Syst., vol. 2017-Decem, no. Nips, pp. 3147–3155, 2017.
[65] O. Gokalp and E. Tasci, "Weighted Voting Based Ensemble Classification with Hyper-parameter Optimization," 2019 Innovations in Intelligent Systems and Applications Conference (ASYU), Izmir, Turkey, 2019, pp. 1-4
[66] L. Shapley, “Notes on the N-person game — II- The value of an N-person game,” Res. Memo., vol. 670, no. 8-21–51, 1951.
[67] J. B. Chen, K. C. Wu, S. H. Moi, L. Y. Chuang, and C. H. Yang, “Deep learning for intradialytic hypotension prediction in hemodialysis patients,” IEEE Access, vol. 8, pp. 82382–82390, 2020.
[68] J. Snoek, H. Larochelle, and R. P. Adams, “Practical Bayesian optimization of machine learning algorithms,” in Advances in Neural Information Processing Systems, Dec 2012, vol. 4, pp. 2951–2959.
[69] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic Minority Over-sampling Technique Nitesh,” Artif. Intell. Res., vol. 16, pp. 321–357, 2002.
[70] Haibo He, Yang Bai, E. A. Garcia and Shutao Li, "ADASYN: Adaptive synthetic sampling approach for imbalanced learning," 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), Hong Kong, 2008, pp. 1322-1328