使用多源信息融合辨識慢性疼痛

Chronic Pain Recognition with multiple sources data fusion

張弘燁(Hung-Ye Chang)

指導教授 ： 林進燈

國立交通大學/光電學院/光電系統研究所/碩士(2019年)

摘要

慢性疼痛疾病不只是台灣所面臨到的問題，也是世界各地所面對的問題，慢性疼痛患者在全球有超過15億人口，由此可知慢性疼痛疾病是非常需要重視的一個議題。在過去的許多研究已經顯示多種的特徵是有辦法去做區別慢性疼痛疾病的，與此同時有些對於臨床疾病的研究也指出不同的特徵融合可以有效的達到互補的效果以利於提高分類的結果，但目前對於慢性疼痛疾病的研究大多還是著重於使用單一種資訊去做判斷。綜合以上論述，本文提出一種基於統計檢定的分類模型，將本文實驗中所得到的唾液、心律變異分析、定量測試實驗、腦波，去做一個不同維度的特徵擷取與融合，我們希望透過這種不同特徵的資訊融合能夠達到資料互補的功用，進而去提高預測慢性疼痛病患的辨識結果。研究結果也顯示，使用多重的特徵去做分類確實有助於提高辨識的結果，且透過我們的模型可以進而再去提高辨識的結果尤其實是在CM和FM病人組的比較中甚至可以分別達到87%和88%的辨識結果。

關鍵字

慢性疼痛 ； 多源信息 ； 機器學習 ； 腦電波 ； 心律變異分析 ； 定量感覺測試 ； 可體松 ； 整合學習

並列摘要

Chronic pain is a common disease not in only Taiwan but also in the world. There are over 1.5 billion patients worldwide. Obviously, chronic pain is an important issue.
In previous study had indicated several features which have ability for patients identification. As the same time , there are also some studies reveal that that multimodal information is benefit for improve the identification about clinical disease. It is believed that multiple sources combination can complement each different source which is benefit for the model training. Based on above research, we interested to know more about the chronic pain field. In this study we collect the EEG , ECG, saliva and do the Quantitative sensory testing information. First of all, we apply machine learning with the coherence features transform from EEG to classify the patients’ group. Second of all we combine coherence with other domain information to optimize the accuracy. In our study we have showed that multiple sources information is advantageous to identify chronic pain patients. The accuracy can reach about 87% and 88% in CM and FM groups comparison.

並列關鍵字

chronic pain ； multi-sources ； machine learning ； electroencephalography ； heart-rate variability ； quantitative sensory testing ； cortisol ； ensemble learning

參考文獻

[1] C. Phillips . “The cost and burden of chronic pain”. Rev Pain. ,2009 Jun;3(1):2–5.

[2] F.M. Blyth, L.M. March, M.K. Nicholas, M.J. “Cousins Chronic pain, work performance and litigation”. Pain 103 2003 41–47.

[3] Z. Cao , CT. Lin , CH. Chuang , KL. Lai , AC. Yang, JL. Fuh,, and SJ. Wang, “Resting-state EEG power and coherence vary between migraine phases”. JOURNAL OF HEADACHE AND PAIN, vol. 17,2016

[4] S. Evans , L. Seidman , J. Tsao , K. Lung , L. Zeltzer and B. Naliboff . “Heart rate variability as a biomarker for autonomic nervous system response differences between children with chronic pain and healthy control children”. Journal of Pain Research. 2013 Jun 12;6:449-57.

[5] K. Blumenstiel , A. Gerhardt ,R. Rolke, C. Bieber , J. Tesarz,H. Friederich H,W. Eich and R. Treede .”Quantitative sensory testing profiles in chronic back pain are distinct from those in fibromyalgia”. The Clinical Journal of Pain. 2011 Oct;27(8):682-90.