文獻(xiàn)解讀|柳葉刀：基于機(jī)器學(xué)習(xí)的急性冠脈綜合征不良事件預(yù)測(cè)：一項(xiàng)匯集數(shù)據(jù)集的建模研究
Background
急性冠脈綜合征(ACS)患者發(fā)生缺血和出血事件的風(fēng)險(xiǎn)很高，兩者都是不良預(yù)后的驅(qū)動(dòng)因素。
風(fēng)險(xiǎn)評(píng)估在每個(gè)患者的臨床管理中起著至關(guān)重要的作用，對(duì)于選擇二級(jí)預(yù)防的最佳藥物治療具有重要意義站蝠。
目前對(duì)急性冠脈綜合征(ACS)后缺血和出血事件的預(yù)測(cè)工具的準(zhǔn)確性對(duì)于個(gè)體化的患者管理策略來(lái)說(shuō)仍然不夠苗沧。
機(jī)器學(xué)習(xí)方法可能能夠克服當(dāng)前分析方法在風(fēng)險(xiǎn)預(yù)測(cè)中的一些限制,且有效性已在幾個(gè)心血管應(yīng)用中得到證明桨踪。
Methods and Result
Datasets
- 為了建立機(jī)器學(xué)習(xí)模型伪节，我們使用了19826名成年急性冠脈綜合征患者(≥18歲)的派生隊(duì)列吼砂，并進(jìn)行了1年的隨訪顷帖。
- 為了評(píng)估模型的性能美旧，我們使用了外部驗(yàn)證隊(duì)列，包括3444名住院的成年ACS患者贬墩，隨訪1年和2年榴嗅。
研究人群的臨床和治療特征
Study outcomes：
特征量選擇
![]()
4.png
結(jié)構(gòu)化數(shù)據(jù)集包括25個(gè)變量:
- 16 clinical variables
- 5 thera- peutic variables
- 2 angiographic variables
- 2 procedural variables
機(jī)器學(xué)習(xí)算法
- K-Nearest Neighbours (KNN)
- Naive Bayes (NB)
- Random Forest (RF)
- Adaptive Boosting (ADB)
機(jī)器學(xué)習(xí)(ML)算法的評(píng)價(jià)
學(xué)習(xí)指標(biāo)
ROC曲線
AUC值
校正圖(觀測(cè)與預(yù)測(cè)風(fēng)險(xiǎn)的十分位數(shù))
-
其他評(píng)價(jià)指標(biāo)如下:
5.png
ROC curves and AUC values
Death, ReAMI and BARC-major bleeding prediction
Performance metrics and algorithm choice for the PRAISE score
Observed vs. Predicted Risk
- Calibration plots
- Observed vs. predicted decile risk comparative bar plots
PRAISE model
AUCs for death, myocardial infarction, and major bleeding for the training, internal validation, and external validation datasets at 1-year follow-up
Risk of observed death, myocardial infarction, and major bleeding according to deciles of event probability based on PRAISE scores
特征相對(duì)重要性
- 在訓(xùn)練過(guò)程中是否選擇變量來(lái)分割節(jié)點(diǎn)中的數(shù)據(jù)
- 平方誤差改善了多少
總之，如果在一個(gè)變量中找到最多的加權(quán)樹(shù)和產(chǎn)生高純度分裂陶舞，它將有較高的相對(duì)重要值嗽测。
scaled importance:每個(gè)變量的相對(duì)重要性與最高的變量相對(duì)重要性之間的比值.
feature importance weight on the PRAISE risk prediction:每個(gè)變量的相對(duì)重要性與所有變量的相對(duì)重要性之和的比值。
Radar plot for the eight most important predictors of death, myocardial infarction, and major bleeding
Classes of risk
stratified by deciles of event prob- ability according to the relating PRAISE score
- low risk: first to sixth deciles;
- intermediate risk: seventh to ninth deciles;
- and high risk: tenth decile
Compared with low risk, being categorised as being at intermediate risk and high risk was associated with increased (p<0·0001) event occurrence for all the PRAISE scores.
Cross-classification of myocardial infarction and major bleeding risk classes and illustration of the hypothetical trade-off between the two types
of risk
PRAISE score with a lower number of patient features
Discussion
- Our score offers very high accuracy in detecting the risk of all-cause death after an ACS in a population treated with current standard therapies.
- According to such stratification, a tenth of patients (the highest decile) would be classified at discharge as being at high risk of either death, recurrent myocardial infarction, or major bleeding, thus being candidates for a tighter follow-up.
Limitations
- The first is the observational retrospective design of the two registries
composing the derivation cohort. - A further possible limitation of our approach can be identified in the slight under estimation of the adaptive boosting classifier among high-risk patients
寫在后面：
小木舟水平有限肿孵，文中難免有些紕漏唠粥，希望各位讀者能夠不吝賜教。歡迎大家關(guān)注我的
B站：木舟筆記停做，獲取更多視頻講解晤愧。制作不易，希望大家多多點(diǎn)贊蛉腌、在看官份。
往期文章
- 跟著CELL學(xué)作圖|1.火山圖
- 跟著Cell學(xué)作圖 | 2.柱狀圖+誤差棒+散點(diǎn)+差異顯著性檢驗(yàn)
- 跟著 Cell 學(xué)作圖 | 3.箱線圖+散點(diǎn)+差異顯著性檢驗(yàn)
- 跟著 Cell 學(xué)作圖 | 4.小提琴圖
- 跟著Cell學(xué)作圖 | 5.UMAP降維分析
- 跟著Cell學(xué)作圖 | 6.時(shí)間序列分析(Mfuzz包)
- 跟著Cell學(xué)作圖|7.富集分析(Metascape數(shù)據(jù)庫(kù))
- 跟著Cell學(xué)作圖|8.富集分析網(wǎng)絡(luò)圖(Cytoscape/ClueGO)
- 跟著Cell學(xué)作圖|9.PPI分析(GeNets數(shù)據(jù)庫(kù))
- 跟著Cell學(xué)作圖|10.復(fù)雜熱圖
- 跟著Cell學(xué)作圖| 11.Ingenuity Pathway Analysis(IPA)
- 跟著Cell學(xué)作圖 | 12.韋恩圖(Vennerable包)
