心智圖資源庫 機器學習演算法與諾莫圖建構對第2型糖尿病患者糖尿病視網膜病變預測
這是一篇關於機器學習演算法和諾莫圖建構對第2型糖尿病患者糖尿病視網膜病變預測的思維導圖,主要內容包括:補充資料,表不如圖,字不如表,摘要,題目。
編輯於2024-10-07 21:30:17In order to help everyone use DeepSeek more efficiently, a collection of DeepSeek guide mind map was specially compiled! This mind map summarizes the main contents: Yitu related links, DS profile analysis, comparison of DeepSeek and ChatGPT technology routes, DeepSeek and Qwen model deployment guide, how to make more money with DeepSeek, how to play DeepSeek, DeepSeek scientific research Application, how to import text from DeepSeek into MindMaster, the official recommendation of DeepSeek Wait, allowing you to quickly grasp the essence of AI interaction. Whether it is content creation, plan planning, code generation, or learning improvement, DeepSeek can help you achieve twice the result with half the effort!
This is a mind map about DeepSeek's 30 feeding-level instructions. The main contents include: professional field enhancement instructions, interaction enhancement instructions, content production instructions, decision support instructions, information processing instructions, and basic instructions.
This is a mind map about a commercial solution for task speech recognition. The main content includes: text file content format:, providing text files according to the same file name as the voice file.
In order to help everyone use DeepSeek more efficiently, a collection of DeepSeek guide mind map was specially compiled! This mind map summarizes the main contents: Yitu related links, DS profile analysis, comparison of DeepSeek and ChatGPT technology routes, DeepSeek and Qwen model deployment guide, how to make more money with DeepSeek, how to play DeepSeek, DeepSeek scientific research Application, how to import text from DeepSeek into MindMaster, the official recommendation of DeepSeek Wait, allowing you to quickly grasp the essence of AI interaction. Whether it is content creation, plan planning, code generation, or learning improvement, DeepSeek can help you achieve twice the result with half the effort!
This is a mind map about DeepSeek's 30 feeding-level instructions. The main contents include: professional field enhancement instructions, interaction enhancement instructions, content production instructions, decision support instructions, information processing instructions, and basic instructions.
This is a mind map about a commercial solution for task speech recognition. The main content includes: text file content format:, providing text files according to the same file name as the voice file.
機器學習演算法與諾莫圖建構對第2型糖尿病患者糖尿病視網膜病變預測的比較
題目
Comparison of Machine Learning Algorithms and Nomogram Construction for Diabetic Retinopathy Prediction in Type 2 Diabetes Mellitus Patients
摘要
背景/目的
糖尿病(diabetes mellitus ,DM)
糖尿病視網膜病變(diabetic retinopathy ,DR)
比較各種機器學習演算法在2型DM患者中建立DR預測模型,並基於最佳模型開發列線圖
方法
這項橫斷面研究包括接受常規DR篩檢的DM患者
將患者隨機分為訓練集( 244例)及驗證集( 105例)
最小絕對收縮和選擇算子回歸用於臨床特徵的選擇
比較了6種機器學習演算法:決策樹( DT )、K近鄰( KNN )、邏輯迴歸模型( LM )、隨機森林( RF )、支援向量機( SVM )和XGBoost ( XGB )
透過受試者工作特徵曲線( ROC )、校準度和決策曲線分析( DCA )評估模型效能。然後在最佳模型的基礎上開發了列線圖
結果
與其他5種機器學習演算法( DT、KNN、RF、SVM、XGB)相比,LM在驗證集中表現出最高的ROC曲線下面積( AUC為0.894)和召回率( 0.92 )
此外,校正曲線和DCA結果也相對較好
LM包括病程、DPN、胰島素用量、尿蛋白、ALB
5個X
列線圖在2個資料集經過1 000次自舉檢驗後均表現出穩健的區分度( AUC :訓練集0.856 ,驗證集0.868)、校準度和臨床適用性
結論
在六種不同的機器學習演算法中,LM演算法表現出了最佳的效能
建立了基於logistic迴歸的預測2型DM患者發生DR的列線圖
此列線圖可作為DR檢測的有價值的工具,便於及時治療
字不如表
Table 1
Logistic模型對DR的預測特性
多因素邏輯迴歸
表不如圖
Figure 1
研究流程圖
Figure 2
透過LASSO回歸模型進行臨床特徵選擇
92個變數
Figure 3
訓練集驗證集多模型ROC、校準曲線、DCA、效能評價
Figure 4
預測2型DM患者發生DR可能性的列線圖
Figure 5
1000次bootstrap的logistic迴歸模型的效能
補充資料
Table S1
訓練集和驗證集中患者的特徵
Figure S1
使用不同的機器學習演算法預測糖尿病視網膜病變的重要性不同