Skip to main content

Table 1 Information of the two datasets based on 3D brain scan, where 9 image slices per 3D scan were used in all our experiments

From: Deep semi-supervised learning for brain tumor classification

Dataset

Tumor

#Patients

#3D scans

#3D scans

#3D scans

#3D scans

 

type

 

(T1/T1ce/

for training

for validation

for testing

   

T2/FLAIR)

(origial/GAN

(original)

(original)

    

augmented)

  

TCGA

IDH mutation

55

55

33/99

6

16

 

IDH wild-type

112

112

66/198

13

33

MICCAI

HGG

210

210

126/126

21

63

 

LGG

75

75

45/45

7

23

  1. Further, the training/validation/testing sets were patient-wise partitioned according to 60%,10%,30% approximately, and 2/3 of the original scans in the training datasets were labeled, and the remaining 1/3 were set as unlabeled
Back to article page

BMC Medical Imaging

ISSN: 1471-2342

Contact us