-
Notifications
You must be signed in to change notification settings - Fork 0
/
preComBat_3img3clin_LowerUpper_050721.R
223 lines (176 loc) · 8.06 KB
/
preComBat_3img3clin_LowerUpper_050721.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
rm(list=ls())
load("dataPreprocessing_CCTSIproject_v2.RData")
library(performanceEstimation)
library(caret)
#Calculate performance accuracy over REP test sets
#Baseline lower bound
#set.seed(20210216)
REP = 75
accuracy.2.lb = matrix(NA, REP, 4)
sensitivityGlio.2.lb = specificityMenin.2.lb = rep(NA, REP)
ind.meninglio.test.CSU = vector(mode = "list", length=REP)
tic = proc.time()
for(r in 1:REP){
print(r)
set.seed(2*r)
#Test data: CSU
menin.CSU.test = sample(nrow(menin.CSU.image3), 19)
glio.CSU.test = sample((nrow(menin.CSU.image3)+1):nrow(meninglio.CSU.image3.1), 19)
ind.meninglio.test.CSU[[r]] = c(menin.CSU.test, glio.CSU.test)
#Scenario 2A: 3 image variables + 3 biological covariates
data.temp.2 = meninglio.Out.image3.2
dim(data.temp.2) #99 x 7
#SMOTE on training data
data.temp.2$dis.lab = factor(ifelse(meninglio.Out.image3.2$dis.lab == 0,"Glio","Menin"))
data.temp.2$sexMF = factor(ifelse(meninglio.Out.image3.2$sexMF == 0,"F","M"))
data.temp.2$breedTypeBO = factor(ifelse(meninglio.Out.image3.2$breedTypeBO == 0,"Non-B","Brachy"))
newData.2 <- smote(dis.lab ~ ., data = data.temp.2, perc.over = 3, perc.under = 1.45)
table(newData.2$dis.lab)
meninglio.Out.image3.2.train = rbind(data.temp.2[which(data.temp.2$dis.lab == "Menin"),],
newData.2[which(newData.2$dis.lab == "Glio"),])
#dim(meninglio.Out.image3.2.train)
#colnames(meninglio.Out.image3.2.train)
meninglio.CSU.image3.2.test = meninglio.CSU.image3.2[ind.meninglio.test.CSU[[r]],]
dim(meninglio.CSU.image3.2.test) #38 x 7
data.temp.CSU.2.test = meninglio.CSU.image3.2.test
data.temp.CSU.2.test$dis.lab = factor(ifelse(meninglio.CSU.image3.2.test$dis.lab == 0,"Glio","Menin"))
data.temp.CSU.2.test$sexMF = factor(ifelse(meninglio.CSU.image3.2.test$sexMF == 0,"F","M"))
data.temp.CSU.2.test$breedTypeBO = factor(ifelse(meninglio.CSU.image3.2.test$breedTypeBO == 0,"Non-B","Brachy"))
dim(data.temp.CSU.2.test) #38 x 7
#Fit logistic model on the training data
train_control = trainControl(method = "repeatedcv",
number = 5,
repeats = 25,
classProbs = TRUE
#seeds = 2*r
#sampling = "up",
#search = "random"
)
#Scenario 2: 3 image variables + 3 biological covariates
mtry <- 1:ncol(meninglio.Out.image3.2.train[,-4])
tunegrid <- expand.grid(.mtry=mtry)
set.seed(2*r)
log.model.Out.train.2 <- train(x = meninglio.Out.image3.2.train[,-4],
y = meninglio.Out.image3.2.train[,4],
#data = meninglio.CSU.image3.1.train,
trControl = train_control,
#preProcess = c("center", "scale"),
method = 'rf',
metric = "Accuracy",
tuneGrid = tunegrid
#family = binomial()
)
#Predictions
#CSU test set
log.predictions.CSU.test.2 <- predict(log.model.Out.train.2, data.temp.CSU.2.test[,-4])
head(log.predictions.CSU.test.2)
#Confusion Matrix
#3 image variables
confMat.test.CSU.2 = confusionMatrix(log.predictions.CSU.test.2, data.temp.CSU.2.test$dis.lab)
#Store the results
#Test set: CSU
accuracy.2.lb[r,] = confMat.test.CSU.2$overall[c(1,3,4,7)]
sensitivityGlio.2.lb[r] = confMat.test.CSU.2$byClass[1]
specificityMenin.2.lb[r] = confMat.test.CSU.2$byClass[2]
}
toc = proc.time() - tic
toc
round(median(sensitivityGlio.2.lb), 3)
round(mad(sensitivityGlio.2.lb), 3)
round(mean(sensitivityGlio.2.lb), 3)
round(sd(sensitivityGlio.2.lb), 3)
round(median(specificityMenin.2.lb), 3)
round(mad(specificityMenin.2.lb), 3)
round(mean(specificityMenin.2.lb), 3)
round(sd(specificityMenin.2.lb), 3)
round(median(accuracy.2.lb[,1]), 3)
round(mad(accuracy.2.lb[,1]), 3)
round(mean(accuracy.2.lb[,1]), 3)
round(sd(accuracy.2.lb[,1]), 3)
median(accuracy.2.lb[,1])
accuracy.2.lb[order(accuracy.2.lb[,1])[38]]
sensitivityGlio.2.lb[order(accuracy.2.lb[,1])[38]]
specificityMenin.2.lb[order(accuracy.2.lb[,1])[38]]
#Calculate performance accuracy over REP test sets
#Baseline UPPER BOUND
#set.seed(20210216)
REP=75
accuracy.2.ub = matrix(NA, REP, 4)
sensitivityGlio.2.ub = specificityMenin.2.ub = rep(NA, REP)
tic = proc.time()
for(r in 1:REP){
print(r)
set.seed(2*r)
#Scenario 2: 3 image variables + 3 clinical covariates
meninglio.CSU.image3.2.train.pre = meninglio.CSU.image3.2[-ind.meninglio.test.CSU[[r]],]
data.temp.2 = meninglio.CSU.image3.2.train.pre[c(sample(87, 79), 88:107), ]
dim(data.temp.2) #99 x 7
#SMOTE on training data
data.temp.2$dis.lab = factor(ifelse(data.temp.2$dis.lab == 0,"Glio","Menin"))
data.temp.2$sexMF = factor(ifelse(data.temp.2$sexMF == 0,"F","M"))
data.temp.2$breedTypeBO = factor(ifelse(data.temp.2$breedTypeBO == 0,"Non-B","Brachy"))
newData.2 <- smote(dis.lab ~ ., data = data.temp.2, perc.over = 3, perc.under = 1.45)
table(newData.2$dis.lab)
meninglio.CSU.image3.2.train = rbind(data.temp.2[which(data.temp.2$dis.lab == "Menin"),],
newData.2[which(newData.2$dis.lab == "Glio"),])
dim(meninglio.CSU.image3.2.train) #159 x 7
meninglio.CSU.image3.2.test = meninglio.CSU.image3.2[ind.meninglio.test.CSU[[r]],]
dim(meninglio.CSU.image3.2.test) #38 x 7
data.temp.CSU.2.test = meninglio.CSU.image3.2.test
data.temp.CSU.2.test$dis.lab = factor(ifelse(meninglio.CSU.image3.2.test$dis.lab == 0,"Glio","Menin"))
data.temp.CSU.2.test$sexMF = factor(ifelse(meninglio.CSU.image3.2.test$sexMF == 0,"F","M"))
data.temp.CSU.2.test$breedTypeBO = factor(ifelse(meninglio.CSU.image3.2.test$breedTypeBO == 0,"Non-B","Brachy"))
dim(data.temp.CSU.2.test) #38 x 7
#Fit logistic model on the training data
train_control = trainControl(method = "repeatedcv",
number = 5,
repeats = 25,
classProbs = TRUE
#sampling = "up",
#search = "random"
)
#Scenario 2: 3 image variables + 3 clinical covariates
mtry <- 1:ncol(meninglio.CSU.image3.2.train[,-4])
tunegrid <- expand.grid(.mtry=mtry)
set.seed(2*r)
log.model.CSU.train.2 <- train(x = meninglio.CSU.image3.2.train[,-4],
y = meninglio.CSU.image3.2.train[,4],
#data = meninglio.CSU.image3.1.train,
trControl = train_control,
#preProcess = c("center", "scale"),
method = 'rf',
metric = "Accuracy",
tuneGrid = tunegrid
#family = binomial()
)
#Predictions
log.predictions.CSU.test.2 <- predict(log.model.CSU.train.2, data.temp.CSU.2.test[,-4])
head(log.predictions.CSU.test.2)
#Confusion Matrix
#Test set: CSU
confMat.test.CSU.2.ub = confusionMatrix(log.predictions.CSU.test.2, data.temp.CSU.2.test$dis.lab)
#Store the results
accuracy.2.ub[r,] = confMat.test.CSU.2.ub$overall[c(1,3,4,7)]
sensitivityGlio.2.ub[r] = confMat.test.CSU.2.ub$byClass[1]
specificityMenin.2.ub[r] = confMat.test.CSU.2.ub$byClass[2]
}
toc = proc.time() - tic
toc
round(median(sensitivityGlio.2.ub), 3)
round(mad(sensitivityGlio.2.ub), 3)
round(mean(sensitivityGlio.2.ub), 3)
round(sd(sensitivityGlio.2.ub), 3)
round(median(specificityMenin.2.ub), 3)
round(mad(specificityMenin.2.ub), 3)
round(mean(specificityMenin.2.ub), 3)
round(sd(specificityMenin.2.ub), 3)
round(median(accuracy.2.ub[,1]), 3)
round(mad(accuracy.2.ub[,1]), 3)
round(mean(accuracy.2.ub[,1]), 3)
round(sd(accuracy.2.ub[,1]), 3)
median(accuracy.2.ub[,1])
accuracy.2.ub[order(accuracy.2.ub[,1])[38]]
sensitivityGlio.2.ub[order(accuracy.2.ub[,1])[38]]
specificityMenin.2.ub[order(accuracy.2.ub[,1])[38]]
#save.image("preComBat_3img3clin_LowerUpper_050721.RData")
#save.image("preComBat_3img3clin_LowerUpper_090121.RData")