臺灣博碩士論文加值系統

目的 : 本研究主要針對接受Q-Switched 1064 nm銣雅鉻雷射治療之女性患者，根據患者臨床資料以及風險因子，採用人工智慧技術分析術後產生發炎後色素沉著併發症(Postinflammatory hyperpigmentation, PIH)之風險因子。
材料與方法 :本研究收集219位從2015年1月至2017年1月期間接受淨膚雷射技術之女性患者資料，以回溯性方式進行評估，並以標準分數(Z-Score)排除了23筆判斷為離群值之患者資料，針對剩餘196位接受淨膚雷射手術患者之資料樣本，以監督式機器學習分類演算法評估風險因子，並使用標準化(Standardization)進行資料前處理優化演算法之預測結果。風險因子共有15項，包括年齡、黑斑、膚色分級、顴骨斑、斑塊、痤瘡、毛孔、雷射劑量、雷射模式、皮表反應、雷射治療次數、彩衝光、果酸換膚、超音波導入、皮膚照護。接著透過最小絕對壓縮挑選法(Least absolute shrinkage and selection operator, LASSO)與隨機森林(Random forest, RF)進行風險因子之重要性排序，逐步建立邏輯斯迴歸(Logistic regression, LR)、隨機森林演算法之分類模型，並搭配階層式迴歸分析(Hierarchical regression)選取預測因子，最後以網格搜尋法(Grid search)進行演算法模型之穩健優化，並根據準確率(Accuracy, ACC)、受試者曲線下面積(Area under the receiver operating characteristic curve, AUC)與陰性預測值(Negative predictive value, NPV)評估演算法準確性，在初步評估模型之準確率後，以LASSO選取之預測因子進行傾向分數配對，進一步探討各項預測因子對於患者接受淨膚雷射手術後發生PIH之因果關係。
結果 : 本研究經由兩種特徵挑選演算法對預測因子進行重要性排序，LASSO挑選的前五項因子分別為痤瘡、皮膚照護、皮表反應、超音波導入、毛孔，其中超音波導入與毛孔對患者發生PIH之勝算比分別為0.62與0.54，呈現為負相關；痤瘡、皮膚照護、皮表反應對患者發生PIH之勝算比分別為6.134、1.186、1.667，呈現為正相關。LASSO挑選之預測因子以及隨機森林挑選之預測因子之演算法，對於以全部風險因子所建立之演算法之優化結果如下：邏輯斯迴歸模型從AUC: 0.708提升為AUC: 0.736；以隨機森林挑選之預測因子建立的隨機森林模型從AUC:0.732提升為AUC:0.781。針對LASSO挑選結果進行網格搜尋之優化結果如下：邏輯斯迴歸模型從AUC: 0.736提升為AUC:0.821；針對隨機森林挑選結果進行網格搜尋之優化結果如下：隨機森林模型從AUC: AUC: 0.781提升為AUC:0.839。
隨機森林挑選之預測因子所建立的穩健化模型之AUC高於以LASSO挑選之預測因子所建立的穩健化模型約0.018。並基於階層式迴歸分析之結果可得知隨機森林在使用皮膚照護、雷射治療次數、年齡、雷射劑量、痤瘡等五項特徵因子進行預測時會達到最大的AUC增益，因此一併將隨機森林挑選之因子與LASSO挑選之因子以傾向分數配對進一步分析這些因子與PIH之關係。經過傾向分數調整後病患資料之特徵分布結果皆非常類似，所有特徵之統計檢定p值皆大於0.05，顯示在統計學上無顯著的差異，再將配對後的八筆資料透過邏輯斯迴歸分析計算勝算比，計算結果為痤瘡、皮膚照護、皮表反應、超音波導入、毛孔、年齡及雷射治療次數分別為16.429、3.732、3.566、2.105、1.748、2.190、2.546，其中年雷射劑量之勝算比皆趨近於1，將針對這八項特徵之相關程度進行近一步的探討其對於PIH之影響。
 
結論 : 本研究所使用的超參數穩健優化大幅提升了演算法分類模型的準確度，我們所提出的LTCP模型能夠準確地判斷接受全臉淨膚雷射治療之患者在術後發生PIH之可能性，可做為臨床醫師判斷患者治療情況之協助工具。實驗結果呈現了使用LASSO和隨機森林挑選出的特徵所建立的模型，在使用網格搜尋法優化超參數後使模型的準確率有了更高的提升，LASSO之AUC: 0.821，隨機森林之AUC: 0.839，二者的AUC評估指標僅相差0.018，且相較於超參數優化前更加具備了穩健性。因此建議臨床醫師在使用銣雅鉻雷射進行臉部淨膚治療時可將本研究之結果做為參考依據，多加注重患者之痤瘡、皮膚照護、皮表反應、超音波導入、毛孔、年齡、雷射治療次數、雷射劑量等八個因子，以降低患者於淨膚雷射手術後發生PIH之風險。
關鍵詞 :
發炎後色素沉著併發症、機器學習、最小絕對壓縮挑選法、邏輯斯迴歸、隨機森林、網格搜尋法、傾向分數配對。

Purpose: To assess the risk factor associated with patients after receiving Q-Switched 1064 nm Nd:YAG (neodymium-doped yttrium aluminum garnet laser treatment for postinflammatory hyperpigmentation complication (PIH) by using artificial intelligence algorithm.
Materials and methods: This retrospective study collected clinical and dosimetric data of 219 female patients who received Q-Switched 1064 nm Nd-Yag laser treatment from January 2015 to January 2017. The 23 outliers which out of3 standard deviations of the mean by using standard score (Z-Score)were excluded. For the remaining data of 196 patients receiving Q-Switched 1064 nm Nd:YAG laser treatment, was used to assess risk factors by using supervised machine learning classification algorithm. Standardization was used for data pre-processing to optimize the prediction results of the machine learning algorithm. Risk factors include age, melasma, skin phototype, zygomatic nevus, lentigines, acne, pores, laser dosage, laser mode, immediately endpoints of laser, treatment frequency, Intense pulsed light, A-hydroxyl acid peels, Sonophoresis, Skin care. Then use Least absolute shrinkage and selection operator (LASSO) and Random forest (RF) to rank the importance of risk factors, and gradually develop the classification of logistic regression (LR) and random forest Model, and use Grid search and Hierarchical regression to select predictors based on Accuracy (ACC), Area under the receiver operating characteristic curve (AUC) and Negative predictive value (NPV) assesses the accuracy of the algorithm. And finally, LASSO is used to identify the influential risk factors for relative score pairing to further explore the causal relationship between the characteristic factors and the PIH of the patient after the skin cleansing laser surgery.
Results: The study uses two feature selection algorithms to rank the importance of predictors. The top five factors selected by LASSO are acne, skin care, skin response, ultrasound, and pores. Among them, ultrasound and pores affect patients. The odds ratios of PIH are 0.596 and 0.64, respectively, showing a negative correlation; The odds ratios of acne, skin care, and skin surface reactions to patients with PIH were 5.75, 1.17, and 2.25, respectively, showing a positive correlation.
The optimization results of the predictor selected by LASSO and the predictor selected by random forest are as follows: Logistic Regression model has been increased from AUC: 0.708 to AUC: 0.736; selected by random forest The random forest model established by the predictor has been increased from AUC: 0.732 to AUC: 0.781. The optimization results of grid search for the LASSO selection results are as follows: the logistic regression model has been increased from AUC: 0.736 to AUC: 0.821; the optimization results of the grid search for the random forest selection results are as follows: the random forest model has been increased from AUC: AUC: 0.781 to AUC: 0.839.The AUC of the robust model established by the predictors selected by random forest is higher than the robust model established by the predictors selected by LASSO by about 0.018. And based on the results of the hierarchical regression analysis, it can be known that the random forest will achieve the largest AUC gain when using five characteristic factors such as skin care, machine treatment times, age, laser dose, and small sores, so it is divided into Random forest selection factors and LASSO selection factors simultaneously analyze the relationship between these factors and PIH with selectivity scores. After the approximate score adjustment, the results of the feature distribution of the patient data are very similar. The statistical test p-values of all features are greater than 0.05, indicating that there is no statistically significant difference. Then the five data after the pairing are analyzed by logistic regression Calculate the odds ratio. The calculated results are Acne, Skin care, Immediately endpoints of laser, Sonophoresis, Pores, Age, TFs are 16.429, 3.732, 3.566, 2.105, 1.748, 2.190and 1.33, respectively, which are positively correlated. We will focus on one of these seven characteristics. The positive and negative correlations will be further explored for their influence on PIH.
Conclusions: The experimental results show that the model built using the features selected by LASSO and random forest, the AUC evaluation index is 3% and 5% higher than the model built using risk factors, respectively. After using the grid search method to optimize the hyperparameters, the accuracy of the model is improved again. The AUC of LASSO: 0.821, and the AUC of random forest: 0.839. The difference between the AUC evaluation indicators of the two is only 0.018, and compared with that before the hyperparameter optimization. More robust. Therefore, it is recommended that clinicians use the results of this study as a reference when using Nd: YAG laser for facial cleansing treatment, and pay more attention to the patient's Acne, Skin care, IEL, Sonophoresis, Pores, Age, TFs, LD to reduce the risk of PIH after the Nd: YAG Laser Toning.
Key words: Postinflammatory hyperpigmentation, machine learning, Least absolute shrinkage and selection operator, Logistic regression, Random forest, Grid search, Propensity score matching.

摘要 .................................................I
ABSTRACT............................................ IV
致謝 ............................................. VIII
目錄 ............................................... IX
表目錄.............................................. XI
圖目錄............................................. XII
符號縮寫.......................................... XIII
符號說明........................................... XIV
第一章 緒論 ..........................................1
1.1 動機.............................................1
1.2 目的.............................................3
1.3 相關文獻探討......................................4
第二章 材料與方法 ................................... 6
2.1 前言.............................................6
2.2 患者特性資料......................................8
2.3 雷射治療技術.....................................10
1.3.1 患者接受臉部雷射淨膚治療之處置...................11
2.4 發炎後色素沉著併發症..............................13
2.5 風險因子.........................................14
2.5.1 風險因子定義....................................15
2.5.2 患者之特徵因子資料分布圖.........................22
2.5.3 淨膚雷射後發生 PIH 之預測因子挑選.................26
2.5.3.1 PIH 預測因子之異常值分析.......................26
2.5.3.2 PIH 預測因子之最小絕對壓縮挑選法分析............27
2.5.3.3 PIH 預測因子之資料挑選流程.....................28
2.6 預測發炎後色素沉著併發症之演算法選擇................30
2.6.1 隨機森林演算法..................................30
2.6.1.1 決策樹演算法..................................31
2.6.1.2 裝袋算法......................................33
2.6.2 邏輯斯迴歸演算法................................33
2.7 發炎後色素沉著併發症之模型效能評估..................36
2.7.1 演算法評估指標..................................36
2.7.1.1 受試者操作特性曲線下面積 AUC 評估..............36
2.8 穩健化發炎後色素沉著併發症機率預測模型..............37
2.9 傾向分數配對......................................38
第三章 結果 ..........................................40
3.1 前言..............................................40
3.2 最小絕對壓縮挑選法之特徵因子排序及挑選結果...........41
3.3 隨機森林之特徵因子排序及挑選結果....................45
3.4 總體演算法模型之 ACC、AUC、NPV 比較.................48
3.5 傾向分數配對之結果.................................50
3.6 預測因子傾向配對後之勝算比..........................60
第四章 討論 ...........................................62
4.1 相關預測因子.......................................62
4.2 監督式演算法模型之分析與比較........................65
4.3 資料預處理方式之分析與比較..........................67
4.4 超參數穩健優化之分析與比較..........................68
4.5 因子變數之分析.....................................69
4.5.1 痤瘡............................................69
4.5.2 皮膚照護........................................69
4.5.3 超音波導入......................................70
4.5.4 皮表反應........................................71
4.5.5 毛孔............................................73
4.5.6 彩衝光..........................................75
4.5.7 顴骨斑..........................................75
4.5.8 年齡............................................75
4.5.9 果酸換膚........................................76
4.5.10 黑斑...........................................76
4.5.11 雷射治療次數....................................77
4.5.12 膚色分級........................................78
4.5.13 斑塊...........................................78
4.5.14 雷射劑量........................................79
4.5.15 雷射模式........................................80
第五章 結論 ...........................................81
參考文獻...............................................82

1.Agbai, O., I. Hamzavi, and J. Jagdeo, Laser treatments for postinflammatory hyperpigmentation: a systematic review. Jama Dermatology, 2017. 153(2): p. 199-206.
2.Pandya, A.G. and I.L. Guevara, Disorders of hyperpigmentation. Dermatologic clinics, 2000. 18(1): p. 91-98.
3.Kaufman, B.P., T. Aman, and A.F. Alexis, Postinflammatory hyperpigmentation: epidemiology, clinical presentation, pathogenesis and treatment. American Journal of Clinical Dermatology, 2018. 19(4): p. 489-503.
4.Abrol, S. and R. Sharma, A clinicoepidemiological study of facial hypermelanosis among females of reproductive age group. Our Dermatology Online/Nasza Dermatologia Online, 2019. 10(3).
5.Arora, P., et al., Lasers for treatment of melasma and post-inflammatory hyperpigmentation. Journal of cutaneous and aesthetic surgery, 2012. 5(2): p. 93.
6.Ye, N., et al., Functional crystals, in Modern Inorganic Synthetic Chemistry. 2017, Elsevier. p. 575-611.
7.Kilmer, S.L., et al., Treatment of epidermal pigmented lesions with the frequency-doubled Q-switched Nd: YAG laser: A controlled, single-impact, dose-response, multicenter trial. Archives of dermatology, 1994. 130(12): p. 1515-1519.
8.FridmanSorokina, A.V., Innovative Methods Of Miscellaneous Skin Inclusions Removal And Post-Inflammatory Hyperpigmentation Correction. European Journal of Molecular & Clinical Medicine, 2020. 7(5): p. 214-221.
9.Topol, E.J., High-performance medicine: the convergence of human and artificial intelligence. Nature medicine, 2019. 25(1): p. 44-56.
10.Bi, W.L., et al., Artificial intelligence in cancer imaging: clinical challenges and applications. CA: a cancer journal for clinicians, 2019. 69(2): p. 127-157.
11.Jamshidi, M., et al., Artificial intelligence and COVID-19: deep learning approaches for diagnosis and treatment. IEEE Access, 2020. 8: p. 109581-109595.
12.Niknejad, A. and D. Petrovic, Introduction to computational intelligence techniques and areas of their applications in medicine. Med Appl Artif Intell, 2013. 51.
13.Won, K.H., et al., A prospective, split-face, double-blinded, randomized study of the efficacy and safety of a fractional 1064-nm Q-switched Nd: YAG laser for photoaging-associated mottled pigmentation in Asian skin. Journal of Cosmetic and Laser Therapy, 2016. 18(7): p. 381-386.
14.Song, H. and H. Kang, Time sequential changes of melanocytes and melanogenic factors in laser-induced postinflammatory hyperpigmentation: P6854. Journal of the American Academy of Dermatology, 2013. 68(4).
15.Silpa-Archa, N., et al., Postinflammatory hyperpigmentation: A comprehensive overview: Epidemiology, pathogenesis, clinical presentation, and noninvasive assessment technique. Journal of the American Academy of Dermatology, 2017. 77(4): p. 591-605.
16.Kang, K.Y., et al., Incidence of cancer among female patients with systemic lupus erythematosus in Korea. Clinical rheumatology, 2010. 29(4): p. 381-388.
17.Panagiotidis, T., T. Stengos, and O. Vravosinos, On the determinants of bitcoin returns: A LASSO approach. Finance Research Letters, 2018. 27: p. 235-240.
18.Mangalathu, S., J.S. Jeon, and R. DesRoches, Critical uncertainty parameters influencing seismic performance of bridges using Lasso regression. Earthquake Engineering Structural Dynamics, 2018. 47(3): p. 784-801.
19.Lee, T.-F., et al., Using multivariate regression model with least absolute shrinkage and selection operator (LASSO) to predict the incidence of xerostomia after intensity-modulated radiotherapy for head and neck cancer. PloS one, 2014. 9(2): p. e89700.
20.Couronné, R., P. Probst, and A.-L. Boulesteix, Random forest versus logistic regression: a large-scale benchmark experiment. BMC bioinformatics, 2018. 19(1): p. 270.
21.Pal, M., Random forest classifier for remote sensing classification. International journal of remote sensing, 2005. 26(1): p. 217-222.
22.Lin, W., et al., An ensemble random forest algorithm for insurance big data analysis. Ieee access, 2017. 5: p. 16568-16575.
23.Manogaran, G. and D. Lopez, Health data analytics using scalable logistic regression with stochastic gradient descent. International Journal of Advanced Intelligence Paradigms, 2018. 10(1-2): p. 118-132.
24.Seem, J.E., Using intelligent data analysis to detect abnormal energy consumption in buildings. Energy and buildings, 2007. 39(1): p. 52-58.
25.Hund, E., D.L. Massart, and J. Smeyers-Verbeke, Inter-laboratory studies in analytical chemistry. Analytica Chimica Acta, 2000. 423(2): p. 145-165.
26.Shiffler, R.E., Maximum Z scores and outliers. The American Statistician, 1988. 42(1): p. 79-80.
27.Colombani, C., et al., Application of Bayesian least absolute shrinkage and selection operator (LASSO) and BayesCπ methods for genomic selection in French Holstein and Montbéliarde breeds. Journal of Dairy Science, 2013. 96(1): p. 575-591.
28.Kong, C., et al., LASSO-based NTCP model for radiation-induced temporal lobe injury developing after intensity-modulated radiotherapy of nasopharyngeal carcinoma. Scientific reports, 2016. 6(1): p. 1-8.
29.Lee, T.-F., et al., LASSO NTCP predictors for the incidence of xerostomia in patients with head and neck squamous cell carcinoma and nasopharyngeal carcinoma. Scientific reports, 2014. 4: p. 6217.
30.Archer, K.J. and R.V. Kimes, Empirical characterization of random forest variable importance measures. Computational statistics & data analysis, 2008. 52(4): p. 2249-2260.
31.Syarif, I., A. Prugel-Bennett, and G. Wills, SVM parameter optimization using grid search and genetic algorithm to improve classification performance. Telkomnika, 2016. 14(4): p. 1502.
32.Fernanda H. Sakamoto, M.M.A.及.R.R.A., Lasers and Other Energy-Based Technologies – Principles and Skin Interactions in Dermatology, Fourth Edition, J.L. Bolognia, MD, Editor. 2018, Elsevier Limited. . p. 2354-2363.
33.Sim, J.H., et al., Treatment of melasma by low-fluence 1064 nm Q-switched Nd: YAG laser. Journal of dermatological treatment, 2014. 25(3): p. 212-217.
34.Silpa-Archa, N., et al., Postinflammatory hyperpigmentation: A comprehensive overview: Epidemiology, pathogenesis, clinical presentation, and noninvasive assessment technique. Journal of the American Academy of Dermatology, 2017. 77(4): p. 591-605.
35.Sugawara, J., et al., Influence of the frequency of laser toning for melasma on occurrence of leukoderma and its early detection by ultraviolet imaging. Lasers in surgery and medicine, 2015. 47(2): p. 161-167.
36.Chan HH; Division of Dermatology, D.o.M., University of Hong Kong, Hong Kong. hhlchan@hkucc.hku.hk, et al., A retrospective analysis of complications in the treatment of nevus of Ota with the Q-switched alexandrite and Q-switched Nd:YAG lasers. Dermatologic Surgery [Dermatol Surg] 2000. Nov; Vol. 26 (11).
37.Chan, N.P., et al., A case series of facial depigmentation associated with low fluence Q‐switched 1,064 nm Nd: YAG laser for skin rejuvenation and melasma. Lasers in surgery and medicine, 2010. 42(8): p. 712-719.
38.Ho, S., et al., A retrospective analysis of the management of acne post‐inflammatory hyperpigmentation using topical treatment, laser treatment, or combination topical and laser treatments in oriental patients. Lasers in surgery and medicine, 2011. 43(1): p. 1-7.
39.William D. James MD, D.M.E.M., James R. Treat MD, Misha A. Rosenbach MD , Isaac M. Neuhaus MD, Disturbances of Pigmentation, in Andrews' Diseases of the Skin, W.D. James, MD; Elston, Dirk M., MD; Treat, James R., MD; Rosenbach, Misha A., MD; Neuhaus, Isaac M., MD, Editor. 2020. p. 862-880.
40.Fitzpatrick, T.B., The validity and practicality of sun-reactive skin types I through VI. Archives of dermatology, 1988. 124(6): p. 869-871.
41.Abad‐Casintahan, F., et al., Frequency and characteristics of acne‐related post‐inflammatory hyperpigmentation. The Journal of dermatology, 2016. 43(7): p. 826-828.
42.Davis, E.C. and V.D. Callender, Postinflammatory hyperpigmentation: a review of the epidemiology, clinical features, and treatment options in skin of color. The Journal of clinical and aesthetic dermatology, 2010. 3(7): p. 20.
43.Cardinali, G., D. Kovacs, and M. Picardo. Mechanisms underlying post-inflammatory hyperpigmentation: lessons from solar lentigo. in Annales de Dermatologie et de Vénéréologie. 2012. Elsevier.
44.Dong, J., J. Lanoue, and G. Goldenberg, Enlarged facial pores: an update on treatments. Cutis, 2016. 98(1): p. 33-6.
45.Henry H.L. Chan , S.G.Y.H., Laser treatment of ethnic skin, in Lasers and Lights: Procedures in Cosmetic Dermatology Series, M. George J. Hruza MD, Elizabeth L. Tanzi MD, FAAD, Jeffrey S. Dover MD, FRCPC, FRCP, Murad Alam MD, MSCI, Editor. 2018. p. 125-144.
46.Wanner, M., et al., Immediate skin responses to laser and light treatments: Therapeutic endpoints: How to obtain efficacy. Journal of the American Academy of Dermatology, 2016. 74(5): p. 821-833.
47.Wong, Y., S.S.J. Lee, and C.L. Goh, Hypopigmentation induced by frequent low-fluence, large-spot-size QS Nd: YAG laser treatments. Annals of dermatology, 2015. 27(6): p. 751-755.
48.Park, J.-H., J.-I. Kim, and W.-S. Kim, Treatment of persistent facial postinflammatory hyperpigmentation with novel pulse-in-pulse mode intense pulsed light. Dermatologic Surgery, 2016. 42(2): p. 218-224.
49.Rossi, A.M. and M.I. Perez, Treatment of hyperpigmentation. Facial Plastic Surgery Clinics, 2011. 19(2): p. 313-324.
50.Grubbs, F.E., Procedures for detecting outlying observations in samples. Technometrics, 1969. 11(1): p. 1-21.
51.Tibshirani, R., Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological), 1996. 58(1): p. 267-288.
52.Lankau, M.J. and T.A. Scandura, An investigation of personal learning in mentoring relationships: Content, antecedents, and consequences. Academy of management Journal, 2002. 45(4): p. 779-790.
53.Bishop, C.M., Pattern recognition and machine learning. 2006: springer.
54.Mitchell, T.M., The discipline of machine learning. Vol. 9. 2006: Carnegie Mellon University, School of Computer Science, Machine Learning ….
55.Witten, I.H. and E. Frank, Data mining: practical machine learning tools and techniques with Java implementations. Acm Sigmod Record, 2002. 31(1): p. 76-77.
56.Breiman, L., Random forests. Machine learning, 2001. 45(1): p. 5-32.
57.Breiman, L., et al., Classification and regression trees. 1984: CRC press.
58.Rokach, L., Ensemble-based classifiers. Artificial intelligence review, 2010. 33(1-2): p. 1-39.
59.Bickel, P.J. and D.A. Freedman, Asymptotic normality and the bootstrap in stratified sampling. The annals of statistics, 1984: p. 470-482.
60.Wichmann, F.A. and N.J. Hill, The psychometric function: II. Bootstrap-based confidence intervals and sampling. Perception psychophysics, 2001. 63(8): p. 1314-1329.
61.Probst, P., M.N. Wright, and A.L. Boulesteix, Hyperparameters and tuning strategies for random forest. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 2019. 9(3): p. e1301.
62.Paper, D. and D. Paper, Scikit-Learn Classifier Tuning from Simple Training Sets. Hands-on Scikit-Learn for Machine Learning Applications: Data Science Fundamentals with Python, 2020: p. 137-163.
63.Rosenbaum, P.R. and D.B. Rubin, The central role of the propensity score in observational studies for causal effects. Biometrika, 1983. 70(1): p. 41-55.
64.Rosenbaum, P.R. and D.B. Rubin, Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. The American Statistician, 1985. 39(1): p. 33-38.
65.Morgan, C.J., Reducing bias using propensity score matching. 2018, Springer.
66.Gayat, E., et al., Propensity scores in intensive care and anaesthesiology literature: a systematic review. Intensive care medicine, 2010. 36(12): p. 1993-2003.
67.Pham, T., et al., Extracorporeal membrane oxygenation for pandemic influenza A (H1N1)–induced acute respiratory distress syndrome: a cohort study and propensity-matched analysis. American journal of respiratory and critical care medicine, 2013. 187(3): p. 276-285.
68.Abad-Casintahan, F., et al., Frequency and characteristics of acne-related post-inflammatory hyperpigmentation. The Journal of Dermatology, 2016. 43(7): p. 826-828.
69.Chen, Y.-T., et al., Combined vitamin C sonophoresis and neodymium-doped yttrium aluminum garnet (NdYAG) laser for facial hyperpigmentation: An outcome observation study in Asian patients. Indian Journal of Dermatology, Venereology, and Leprology, 2016. 82(5): p. 587.
70.Kang, H.J., et al., Postinflammatory hyperpigmentation associated with treatment of solar lentigines using a Q-Switched 532-nm Nd: YAG laser: a multicenter survey. Journal of Dermatological Treatment, 2017. 28(5): p. 447-451.
71.van der Schaaf, A., et al., Multivariate modeling of complications with data driven variable selection: guarding against overfitting and effects of data set size. 2012. 105(1): p. 115-121.
72.Li, W., et al., Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging. Journal of biomedical optics, 2015. 20(12): p. 121305.
73.Eitrich, T. and B. Lang, Efficient optimization of support vector machine learning parameters for unbalanced datasets. Journal of computational and applied mathematics, 2006. 196(2): p. 425-436.
74.Liu, C.-H., et al., Evaluating the Risk Factors of Post Inflammatory Hyperpigmentation Complications with Nd-YAG Laser Toning Using LASSO-Based Algorithm. Applied Sciences, 2020. 10(6): p. 2049.
75.Taylor, M.B., et al., Successful short-term and long-term treatment of melasma and postinflammatory hyperpigmentation using vitamin C with a full-face iontophoresis mask and a mandelic/malic acid skin care regimen. Journal of drugs in dermatology: JDD, 2013. 12(1): p. 45.
76.Se Young Na, K.C.P., Assessment and Treatment of Photoaging-related Mottled Pigmentation, in Surgery of the Skin E-Book: Procedural Dermatology, J.K. Robinson, et al., Editors. 2014, Elsevier Health Sciences.
77.Molinar, V.E., S.C. Taylor, and A.G. Pandya, What’s new in objective assessment and treatment of facial hyperpigmentation? Dermatologic clinics, 2014. 32(2): p. 123-135.
78.Smeltzer, W.L., COSMECEUTICALS AND SKIN CARE, in Pfenninger and Fowler's Procedures for Primary Care E-Book, G.C. Fowler, Editor. 2019, Elsevier Health Sciences.
79.Lee, M.-C., et al., Treatment of melasma with mixed parameters of 1,064-nm Q-switched Nd: YAG laser toning and an enhanced effect of ultrasonic application of vitamin C: a split-face study. Lasers in medical science, 2015. 30(1): p. 159-163.
80.Won, K.H., M.H. Lee, and S.E. Chang, P259: A prospective, split-face, randomized study of the efficacy and safety of a fractional 1064-nm Q-switched Nd: YAG laser for melasma and photoaging. 프로그램북 (구 초록집), 2015. 67(2): p. 509-510.
81.Roh, M.R., H.J. Chung, and K.Y. Chung, Effects of various parameters of the 1064 nm Nd: YAG laser for the treatment of enlarged facial pores. Journal of dermatological treatment, 2009. 20(4): p. 223-228.
82.Kim, B., et al., Sebum, acne, skin elasticity, and gender difference–which is the major influencing factor for facial pores? Skin Research and Technology, 2013. 19(1): p. e45-e53.
83.Jimenez, F., E. Poblet, and A. Izeta, Reflections on how wound healing‐promoting effects of the hair follicle can be translated into clinical practice. Experimental Dermatology, 2015. 24(2): p. 91-94.
84.Chou, W.C., et al., Direct migration of follicular melanocyte stem cells to the epidermis after wounding or UVB irradiation is dependent on Mc1r signaling. Nature medicine, 2013. 19(7): p. 924-929.
85.Mathew, M.L., et al., Intense pulsed light therapy for acne-induced post-inflammatory erythema. Indian dermatology online journal, 2018. 9(3): p. 159.
86.Leong Ee, H., et al., Treatment of acquired bilateral nevus of ota‐like macules (Hori's nevus) with a combination of the 532 nm Q‐Switched Nd: YAG laser followed by the 1,064 nm Q‐switched Nd: YAG is more effective: prospective study. Dermatologic surgery, 2006. 32(1): p. 34-40.
87.Park, J.M., H. Tsao, and S. Tsao, Combined use of intense pulsed light and Q‐switched ruby laser for complex dyspigmentation among Asian patients. Lasers in Surgery and Medicine: The Official Journal of the American Society for Laser Medicine and Surgery, 2008. 40(2): p. 128-133.
88.Grover, C. and B. Reddu, The therapeutic value of glycolic acid peels in dermatology. Indian Journal of Dermatology, Venereology, and Leprology, 2003. 69(2): p. 148.
89.Kim, J., et al., Histopathological study of the treatment of melasma lesions using a low‐fluence Q‐switched 1064‐nm neodymium: yttrium–aluminium–garnet laser. Clinical and Experimental Dermatology: Clinical dermatology, 2013. 38(2): p. 167-171.
90.Tian, B., Laser toning for melasma: A single-centre experience with 38 970 cases. Journal of Cosmetic and Laser Therapy, 2017. 19(3): p. 140-142.
91.Choi, C.P., et al., Retrospective analysis of melasma treatment using a dual mode of low-fluence Q-switched and long-pulse Nd: YAG laser vs. low-fluence Q-switched Nd: YAG laser monotherapy. Journal of Cosmetic and Laser Therapy, 2015. 17(1): p. 2-8.
92.Matsunaga, K.N.H.A.S.T.Y.Y.S.W.K., Comparative study of treatment efficacy and the incidence of post‐inflammatory hyperpigmentation with different degrees of irradiation using two different quality‐switched lasers for removing solar lentigines on Asian skin. JEDV. 27(3): p. 307-312.
93.Wanner, M., et al., Immediate skin responses to laser and light treatments: warning endpoints: how to avoid side effects. Journal of the American Academy of Dermatology, 2016. 74(5): p. 807-819