本文選題：帕金森用藥推薦　切入點：主成分分析　出處：《南京大學(xué)》2014年碩士論文　論文類型：學(xué)位論文

【摘要】：帕金森病是一種常見于中老年的神經(jīng)變性疾病,但隨著人們生活的多樣化,中年以下的人群也有可能患帕金森病。迄今為止,帕金森病是一種沒有辦法治愈并且病情逐步加重的疾病,但是良好的藥物和必要的治療手段可以幫助帕金森病人恢復(fù)機能。目前,帕金森病診斷最重要的依據(jù)來自于帕金森病量表的測評結(jié)果。基于此,研究一種基于量表測評的自動識別帕金森病和帕金森用藥推薦的方法對于預(yù)防、治療帕金森具有非常重要的意義。 本文采用機器學(xué)習(xí)技術(shù)對帕金森用藥推薦模型進行了研究,主要工作如下： 1)提出一種基于支持向量機的自動識別帕金森病的模型,該模型主要是基于一張設(shè)計的新量表快速識別帕金森病,而新量表則是基于主成分分析的帕金森病量表優(yōu)化算法設(shè)計的。實驗發(fā)現(xiàn),用僅占原西醫(yī)量表總問題數(shù)21%的新量表能達(dá)到與原量表相當(dāng)?shù)呐两鹕∽R別水平。 2)針對帕金森用藥推薦模型,結(jié)合了“k標(biāo)簽子集準(zhǔn)則”和“k近鄰準(zhǔn)則”,提出了基于混合策略的多標(biāo)簽學(xué)習(xí)框架用于帕金森病藥物推薦。實驗發(fā)現(xiàn),該算法在帕金森病量表-用藥數(shù)據(jù)集上比ML-kNN具有更好的性能。 3)基于上述工作,設(shè)計并實現(xiàn)了基于“云平臺+終端”的帕金森綜合診療平臺,分別開發(fā)了該平臺的服務(wù)器系統(tǒng)和Android客戶端。
[Abstract]:Parkinson's disease is a common neurodegenerative disease in the middle and old age, but with the diversity of people's lives, people under middle age are also likely to suffer from Parkinson's disease. Parkinson's disease is a disease that can't be cured and is getting worse, but good drugs and necessary treatments can help Parkinson's patients recover. The most important basis for the diagnosis of Parkinson's disease comes from the results of the Parkinson's disease scale. The treatment of Parkinson's disease is of great significance. In this paper, the machine learning technology is used to study the Parkinson's drug recommendation model. The main work is as follows:. 1) A model of automatic recognition of Parkinson's disease based on support vector machine (SVM) is proposed. The model is mainly based on a new scale designed to quickly identify Parkinson's disease. The new scale was designed based on the principle component analysis (PCA) algorithm of Parkinson's disease scale. It was found that the new scale, which only accounted for the total number of problems in the original Western medicine scale (21%), could achieve the same level of Parkinson's disease recognition as the original scale. 2) according to Parkinson's drug recommendation model, combining "k-label subset criterion" and "k-nearest neighbor criterion", a multi-label learning framework based on mixed strategy is proposed for drug recommendation of Parkinson's disease. The algorithm has better performance than ML-kNN in Parkinson's disease scale-drug data set. 3) based on the above work, the Parkinsonian integrated diagnosis and treatment platform based on "cloud platform terminal" is designed and implemented, and the server system and Android client of the platform are developed respectively.
【學(xué)位授予單位】：南京大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：R742.5;TP18

【參考文獻】

相關(guān)期刊論文 前3條

1 董青,李建萍,劉建軍,李焰生;帕金森病患者紋狀體多巴胺轉(zhuǎn)運體顯像與帕金森病臨床量表評分的相關(guān)性[J];臨床神經(jīng)病學(xué)雜志;2005年03期

2 高曉蘭,王文志,胡長梅,楊卿,龍潔,李藝影,趙淑清,周寶玉,趙秀華,徐忠寶,李金,陳清棠;腦卒中數(shù)據(jù)庫研究的方法學(xué)及基線資料分析[J];中華神經(jīng)外科雜志;1999年01期

3 張媛;劉新峰;徐格林;;南京卒中注冊項目的初步研究結(jié)果[J];中國卒中雜志;2006年01期

，

本文編號：1586384