How to get a "taste" of an experienced surgeon or the predictive model of a young graduate student

Ultrasound examination of the arteries in the wrist.  The arrow indicates the radial artery, which has no blood flow because it closed after endovascular heart surgery.
Ultrasound examination of the arteries in the wrist. The arrow indicates the radial artery, which has no blood flow because it closed after endovascular heart surgery.

Introduction





10- - , “, ”. . , . , . , . , . 





, — - - . . ( - ) - ( ) ( ) - - . . - , . 





?

. — : , , . , . , , , . , - - . , ( ). , .  , - , . - , . . , , . 





, ( ), . , — , ( 6 French - 2 ). - , , , - . - , , - , , - , . , - . , 30 . , -  .





, , . , 3 (, ), 10 ( — ).





— , , - «» «». 





«» — , - - 8,8%. , «» , , , 3 %. , «» 988 , 90%, α 1%, 5% «» «». , , 1000 — . . 





. — . , , . 





R ?!

, . , , . , (SRSS STATISTICA) . , , , . Rstudio. — , , , STATISTICA . 





, , . , «» , «» , . — . 





,- - R?! - )). 





( , 0 — , 1 -), ( 25, , , , ), , - .





glm( LO ~ ., data = Final1, family = 'binomial')
      
      



LO — «late occlusion», «Final1». «fit»





fit<-glm( LO ~., Final1, family = 'binomial')
      
      



. R «step» , , — . , ( «Anova»). - ! 





step(fit,direction = 'backward')
      
      



, , — 2 , ( 1,2 ) 





, . «Prob». , , . «Prob» — . 





Final1$Prob<-predict(object=fit,type = «response»)
      
      



  ROC , «ROCR» «predict» 





, 0,75, .





pred_fit<-prediction(Final1$Prob,Final1$LO)
perf_fit<-performance(pred_fit,«tpr»,«fpr»)
plot(perf_fit,colorize=T, print.cutoffs.at=seq(0,1,by=0.1))
auc<-performance(pred_fit,measure = "auc")
str(auc)
      
      



 





perf6<-performance(pred_fit,x.measure = "cutoff",measure = "spec")
plot(perf6,col="blue",lwd=2)
perf7<-performance(pred_fit,x.measure = "cutoff",measure = "sens")
plot(add=T,perf7,col="green",lwd=2)
perf8<-performance(pred_fit,x.measure = "cutoff",measure = "acc")
plot(add=T,perf8,col="black",lwd=2)
abline(v = 0.085,lwd=2)
      
      



, ( )





pred_resp, : . . 0,085, , , .





Final1$pred_resp<-factor(ifelse(Final1$prob>0.085,1,0), labels = c("0","1"))
      
      



“mean” . 70,7%, .





> Final1$correct<-ifelse(Final1$pred_resp==Final1$LO,1,0)
> mean(Final1$correct)
      
      



. , , , . , , , . , , , . 





, , R. . , “/ ” “ ”. , . , , , . - ( - ) ( INTERIN Promis). , 2 - “, 3 -  ” - “” , .








All Articles