根據聯合國 2019 年對全球 65 歲以上老年人的調查，65 歲以上老年人的比例正
在增加，甚至未來的估計也沒有顯示飽和或停止的趨勢。根據台灣國健署的統計，台
灣 65 歲以上老年人的肌少症 (Sarcopenia)患病率為 21.1％，男性為 23.6％，女性為
18.6％。換句話說，近五分之一的人有罹患肌少症和高度殘疾的風險。但是，目前對
肌少症的診斷過程極其複雜且不便，特別是在測量骨骼肌肌肉量（肌少症診斷中的最
後一步）時，這意味著民眾對自己的肌肉狀態還不夠了解以及難以預防民眾罹患肌少
症。
因此，我們利用無痛、非侵入性、易於操作和可攜帶式的電阻抗肌電圖（Electrical
Impedance Myography, EIM）系統和機器學習來建構僅使用指定的肌肉阻抗即可估算
骨骼肌肉質量的模型。在這項研究中，我們使用六種算法來建立用於估計肌肉質量的
回歸模型，例如 Ridge，支持向量回歸（SVR），決策樹，隨機森林（RF），梯度提
升決策樹（GBDT）和極限梯度提升（XGBoost）。數據集包括 96 個樣本和 37 個特
徵。 然後使用交叉驗證來評估模型的性能和進行微調。 測試數據集中的 Ridge 、
SVR、決策樹、隨機森林、GBDT 和 XGBoost 的最終 R2 分別為 0.80、0.929、0.795、0.841、0.898 和 0.910。 再者，測試數據集中 Ridge，SVR，決策樹，隨機森林，GBDT和 XGBoost 的最終 RMSE 分別為 1.432、0.980、2.384、1.454、1.344 和 1.121 公斤。它表明 EIM 特定部位肌肉的原始數據可以用來估計全身骨骼肌質量，而這些肌肉可以為臨床人員提供另一有價值的骨骼肌狀況的信息。

According to the 2019 United Nations survey of the elderly global population, the
proportion of people exceeding 65 years of age is increasing. Moreover, future estimates do not indicate this trend to saturate or drop. Based on the statistics of the National Health Administration, the prevalence rate of sarcopenia in Taiwan among the elderly over 65 years of age is 21.1% (23.6% among men and 18.6% among women). In other words, approximately one of every five elderly people are at risk of sarcopenia and severe disability. However, the current diagnostic process for sarcopenia is extremely complicated and inconvenient, especially in the measurement of skeletal muscle mass, which is the final step in diagnosing this disease. This means that typically, people are not sufficiently aware of their muscle state, and the prevention of sarcopenia among the public in general is limited.
Accordingly, a painless, non-invasive, easy to operate, and portable electrical
impedance myography (EIM) system is developed with machine learning to build a model for estimating skeletal muscle mass based on the impedance of a specific muscle. Six algorithms were employed to build regression models for estimating muscle mass: ridge regression, support vector regression (SVR), decision tree, random forest, gradient boosting decision tree (GBDT), and extreme gradient boosting (XGBoost). There were 96 samples and 37 features in the dataset used. To evaluate and fine-tune the performance of the models, fivefold cross-validation was applied. Using the testing dataset, the final R2 values of ridge regression, SVR, decision tree, random forest, GBDT, and XGBoost were 0.80, 0.929, 0.795, 0.841, 0.898, and 0.910, respectively; the final root mean square errors were 1.432, 0.980,
2.384, 1.454, 1.344, and 1.121 kg, respectively. The foregoing indicates that the EIM raw data of site-specified muscle may be utilized to estimate the skeletal muscle mass. These muscles can provide valuable insight into the state of the skeletal muscle, thus aiding the clinical staff in conducting examinations.

[1] G. McNicoll, “United Nations Department Of Economic and Social Affairs, Population Division: Population, Resources, Environment and Development
Database, Version 4.0,” Popul. Dev. Rev., vol. 32, no. 4, pp. 790–791, Dec. 2006.
[2] C.-J. Yeh, H.-Y. Chang, and W.-H. Pan, “Time trend of obesity, the metabolic
syndrome and related dietary pattern in Taiwan: from NAHSIT 1993-1996 to
NAHSIT 2005-2008,” Asia Pac. J. Clin. Nutr., vol. 20, no. 2, pp. 292–300, 2011.
[3] T. N. Kim and K. M. Choi, “Sarcopenia: Definition, Epidemiology, and
Pathophysiology,” J. Bone Metab., vol. 20, no. 1, pp. 1–10, May 2013.
[4] R. D. Boutin, L. Yao, R. J. Canter, and L. Lenchik, “Sarcopenia: Current Concepts
and Imaging Implications,” Am. J. Roentgenol., vol. 205, no. 3, pp. W255–W266,
Sep. 2015.
[5] H. G. J. Kortman, S. C. Wilder, T. R. Geisbush, P. Narayanaswami, and S. B.
Rutkove, “Age- and gender-associated differences in electrical impedance values of skeletal muscle,” Physiol. Meas., vol. 34, no. 12, pp. 1611–1622, Dec. 2013.
[6] T. Abe, J. P. Loenneke, R. S. Thiebaud, and T. Fukunaga, “Age-related site-specific muscle wasting of upper and lower extremities and trunk in Japanese men and women,” Age, vol. 36, no. 2, pp. 813–821, Apr. 2014.
[7] T. Abe et al., “Age-related, site-specific muscle loss in 1507 Japanese men and
women aged 20 to 95 years,” p. 6, Mar. 2011.
[8] L.-K. Chen et al., “Asian Working Group for Sarcopenia: 2019 Consensus Update on Sarcopenia Diagnosis and Treatment,” J. Am. Med. Dir. Assoc., vol. 21, no. 3, pp. 300-307.e2, Mar. 2020.
[9] S. Strandberg, M.-L. Wretling, T. Wredmark, and A. Shalabi, “Reliability of
computed tomography measurements in assessment of thigh muscle cross-sectional area and attenuation,” BMC Med. Imaging, vol. 10, p. 18, Aug. 2010.
[10] S. B. Heymsfield, Z. Wang, R. N. Baumgartner, and R. Ross, “Human body
composition: advances in models and methods,” Annu. Rev. Nutr., vol. 17, pp. 527– 558, 1997.73
[11] R. Ross, J. Rissanen, H. Pedwell, J. Clifford, and P. Shragge, “Influence of diet and exercise on skeletal muscle and visceral adipose tissue in men,” J. Appl. Physiol. Bethesda Md 1985, vol. 81, no. 6, pp. 2445–2455, Dec. 1996.
[12] M. C. Erlandson, A. L. Lorbergs, S. Mathur, and A. M. Cheung, “Muscle analysis
using pQCT, DXA and MRI,” Eur. J. Radiol., vol. 85, no. 8, pp. 1505–1511, Aug.
2016.
[13] G. M. Blake and I. Fogelman, “Technical principles of dual energy x-ray
absorptiometry,” Semin. Nucl. Med., vol. 27, no. 3, pp. 210–228, Jul. 1997.
[14] F. Buckinx et al., “Pitfalls in the measurement of muscle mass: a need for a reference standard: Measurement of muscle mass,” J. Cachexia Sarcopenia Muscle, vol. 9, no. 2, pp. 269–278, Apr. 2018.
[15] S. Khalil, M. Mohktar, and F. Ibrahim, “The Theory and Fundamentals of
Bioimpedance Analysis in Clinical Status Monitoring and Diagnosis of Diseases,”
Sensors, vol. 14, no. 6, pp. 10895–10928, Jun. 2014.
[16] S. B. Rutkove, “Electrical impedance myography: Background, current state, and future directions,” Muscle Nerve, vol. 40, no. 6, pp. 936–946, Dec. 2009.
[17] B. Sanchez and S. B. Rutkove, “Electrical Impedance Myography and Its
Applications in Neuromuscular Disorders,” Neurotherapeutics, vol. 14, no. 1, pp.
107–118, Jan. 2017.
[18] S. B. Rutkove et al., “Electrical impedance myography as a biomarker to assess ALS progression,” Amyotroph. Lateral Scler. Off. Publ. World Fed. Neurol. Res. Group Mot. Neuron Dis., vol. 13, no. 5, pp. 439–445, Sep. 2012.
[19] S. B. Rutkove et al., “Electrical impedance myography for assessment of Duchenne muscular dystrophy: EIM in DMD,” Ann. Neurol., vol. 81, no. 5, pp. 622–632, May 2017.
[20] R. Aaron, G. J. Esper, C. A. Shiffman, K. Bradonjic, K. S. Lee, and S. B. Rutkove,
“Effects of age on muscle as measured by electrical impedance myography,” Physiol. Meas., vol. 27, no. 10, pp. 953–959, Oct. 2006.
[21] A. W. Tarulli et al., “Electrical Impedance Myography in the Assessment of Disuse Atrophy,” Arch. Phys. Med. Rehabil., vol. 90, no. 10, pp. 1806–1810, Oct. 2009.
[22] Z. Li et al., “Handheld Electrical Impedance Myography Probe for Assessing Carpal Tunnel Syndrome,” Ann. Biomed. Eng., 2017.74
[23] P. Narayanaswami, A. J. Spieker, P. Mongiovi, J. C. Keel, S. C. Muzin, and S. B.
Rutkove, “Utilizing a handheld electrode array for localized muscle impedance
measurements,” Muscle Nerve, vol. 46, no. 2, pp. 257–263, 2012.
[24] S. B. Rutkove, A. Pacheck, and B. Sanchez, “Sensitivity distribution simulations of surface electrode configurations for electrical impedance myography,” Muscle
Nerve, vol. 56, no. 5, pp. 887–895, Nov. 2017.
[25] M. Jafarpoor, J. Li, J. K. White, and S. B. Rutkove, “Optimizing Electrode
Configuration for Electrical Impedance Measurements of Muscle via the Finite
Element Method,” IEEE Trans. Biomed. Eng., vol. 60, no. 5, pp. 1446–1452, May
2013.
[26] S. B. Rutkove, K. S. Lee, C. A. Shiffman, and R. Aaron, “Test-retest reproducibility of 50 kHz linear-electrical impedance myography,” Clin. Neurophysiol. Off. J. Int. Fed. Clin. Neurophysiol., vol. 117, no. 6, pp. 1244–1248, Jun. 2006.
[27] B. Sanchez, A. Pacheck, and S. B. Rutkove, “Guidelines to electrode positioning for human and animal electrical impedance myography research,” Sci. Rep., vol. 6, no. 1, Dec. 2016.
[28] W. R. Frontera et al., “Muscle fiber size and function in elderly humans: a
longitudinal study,” J. Appl. Physiol., vol. 105, no. 2, pp. 637–642, Aug. 2008.
[29] W. You, Z. Yang, and G. Ji, “PLS-based recursive feature elimination for highdimensional small sample,” Knowl.-Based Syst., vol. 55, pp. 15–28, Jan. 2014.
[30] S. Maldonado and R. Weber, “A wrapper method for feature selection using Support Vector Machines,” Inf. Sci., vol. 179, no. 13, pp. 2208–2217, Jun. 2009.
[31] A. Altmann, L. Toloşi, O. Sander, and T. Lengauer, “Permutation importance: a
corrected feature importance measure,” Bioinformatics, vol. 26, no. 10, pp. 1340–
1347, May 2010.
[32] A. Fisher, C. Rudin, and F. Dominici, “All Models are Wrong, but Many are Useful: Learning a Variable’s Importance by Studying an Entire Class of Prediction Models Simultaneously,” ArXiv180101489 Stat, Dec. 2019.
[33] T. Evgeniou and M. Pontil, “Support Vector Machines: Theory and Applications, vol. 2049, pp. 249–257, Jan. 2001.
[34] C.-W. Hsu, C.-C. Chang, and C.-J. Lin, “A Practical Guide to Support Vector
Classification,” p. 16.75
[35] T. Hastie, “Ridge Regularizaton: an Essential Concept in Data Science,”
ArXiv200600371 Cs Stat, May 2020.
[36] G. Guo, H. Wang, D. Bell, and Y. Bi, “KNN Model-Based Approach in
Classification,” Aug. 2004.
[37] L. Rokach and O. Maimon, “Decision Trees,” in The Data Mining and Knowledge Discovery Handbook, vol. 6, 2005, pp. 165–192.
[38] L. Breiman, “Random Forests,” Mach. Learn., vol. 45, no. 1, pp. 5–32, Oct. 2001.
[39] T. Chengsheng, L. Huacheng, and X. Bing, “AdaBoost typical Algorithm and its
application research,” MATEC Web Conf., vol. 139, p. 00222, Jan. 2017.
[40] Y. Shi, J. Li, and Z. Li, “Gradient Boosting With Piece-Wise Linear Regression
Trees,” ArXiv180205640 Cs, Jun. 2019.
[41] S. G. C. G and B. S. -, “Grid Search Tuning of Hyperparameters in Random Forest Classifier for Customer Feedback Sentiment Prediction,” Int. J. Adv. Comput. Sci. Appl., vol. 11, no. 9, 2020.
[42] T. Abe, R. S. Thiebaud, J. P. Loenneke, M. Loftin, and T. Fukunaga, “Prevalence of site-specific thigh sarcopenia in Japanese men and women,” Age, vol. 36, no. 1, pp. 417–426, Feb. 2014