update medmnist+ details

on_medmnist_plus.md

@@ -9,42 +9,42 @@ We will describe the details of MedMNIST+, compared with MedMNIST. More details
 ### PathMNIST
 In MedMNIST, we resize the source images of 3×224×224 into 3×28×28. 
 
-In MedMNIST+, for size 224x224, we utilize the source images directly; while for size 64x64 and 128x128, we resize the source images into the target size. 
+In MedMNIST+, for size 224x224, we utilize the source images directly; while for size 64x64 and 128x128, we resize the source images of 3×224×224 into the target size. 
 
 ### ChestMNIST
 In MedMNIST, we resize the source images of 1×1,024×1,024 into 1×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we resize the source images into the target size. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we resize the source images of 1×1,024×1,024 into the target size. 
 
 ### DermaMNIST
 In MedMNIST, the source images of 3×600×450 are resized into 3×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we resize the source images into the target size. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we resize the source images of 3×600×450 into the target size. 
 
 ### OCTMNIST
 In MedMNIST, the source images are gray-scale, and their sizes are (384–1,536)×(277–512). We center-crop the images with a window size of length of the short edge and resize them into 1×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we center-crop the images with a window size of length of the short edge, and then resize the source images into the target size. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we center-crop the images of (384–1,536)×(277–512) with a window size of length of the short edge, and then resize the source images into the target size. 
 
 ### PneumoniaMNIST
 In MedMNIST, the source images are gray-scale, and their sizes are (384–2,916)×(127–2,713). We center-crop the images with a window size of length of the short edge and resize them into 1×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we center-crop the images with a window size of length of the short edge, and then resize the source images into the target size. Note that after center-crop, the smallest size of the images is 127x127, and we upsample them to 128x128 and 224x224. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we center-crop the source images of (384–2,916)×(127–2,713) with a window size of length of the short edge, and then resize the source images into the target size. Note that after center-crop, the smallest size of the images is 127x127, and we upsample them to 128x128 and 224x224. 
 
 ### RetinaMNIST
 In MedMNIST, the source images of 3×1,736×1,824 are center-cropped with a window size of length of the short edge and resized into 3×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we first center-crop the images with a window size of length of the short edge, and then resize the source images into the target size. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we first center-crop the source images of 3×1,736×1,824 with a window size of length of the short edge, and then resize the source images into the target size. 
 
 ### BreastMNIST
 In MedMNIST, the source images of 1×500×500 are resized into 1×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we resize the source images into the target size. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we resize the source images of 1×500×500 into the target size. 
 
 ### BloodMNIST
 In MedMNIST, the source images with resolution 3×360×363 pixels are center-cropped into 3×200×200, and then resized into 3×28×28. 
 
-In MedMNIST+, for size 224x224, 128x128, and 64x64, we first center-crop the images into 3×200×200, and then resize the source images into the target size. 
+In MedMNIST+, for size 224x224, 128x128, and 64x64, we first center-crop the images with resolution 3×360×363 into 3×200×200, and then resize the source images into the target size. 
 
 ### TissueMNIST
 In MedMNIST, each gray-scale image is 32×32×7 pixels, where 7 denotes 7 slices. We obtain 2D maximum projections by taking the maximum pixel value along the axial-axis of each pixel, and resize them into 28×28 gray-scale images. 
@@ -57,20 +57,19 @@ In MedMNIST, Hounsfeld-Unit (HU) of the 3D images are transformed into gray-scal
 In MedMNIST+, following the same procedure in MedMNIST, we crop 2D images from the center slices of the 3D bounding boxes in axial/coronal/sagittal views (planes). Then, for size 224x224, 128x128, and 64x64, we resize the source images into the target size. 
 
 ### OrganMNIST3D
-In MedMNIST, the source of the OrganMNIST3D is the same as that of the Organ{A,C,S}MNIST. Instead of 2D images, we directly use the 3D bounding boxes and process the images into 28×28×28 to perform multi-class classifcation of 11 body organs. 
+In MedMNIST, the source of the OrganMNIST3D is the same as that of the Organ{A,C,S}MNIST. Instead of 2D images, we directly use the 3D bounding boxes of size (15-152)x(15-112)x(12-1145) and process the images into 28×28×28 to perform multi-class classifcation of 11 body organs. 
 
-In MedMNIST+, we directly use the 3D bounding boxes and process the images into 64x64x64. 
+In MedMNIST+, we directly use the 3D bounding boxes of size (15-152)x(15-112)x(12-1145) and process the images into 64x64x64. 
 
 ### NoduleMNIST3D
-In MedMNIST, we center-crop the spatially normalized images (with a spacing of 1mm × 1mm x 1mm) into 28×28×28. 
+In MedMNIST, we center-crop the spatially normalized images of size 80x80x80 into 28×28×28. 
 
-In MedMNIST+, to be consistent with the dataset in MedMNIST, we center-crop the spatially normalized images (with a spacing of 1mm × 1mm x 1mm) into 28×28×28, and then upsample them into 64x64x64.
+In MedMNIST+, to be consistent with the dataset in MedMNIST, we center-crop the spatially normalized images of size 80x80x80 into 28×28×28, and then upsample them into 64x64x64.
 
 ### AdrenalMNIST3D
-In MedMNIST, we calculate the center of adrenal and resize the center-cropped
-64×64×64 volume into 28×28×28. 
+In MedMNIST, we center-crop the images of size 120x120x120 into 64×64×64, and then resize them into 28×28×28. 
 
-In MedMNIST+, we calculate the center of adrenal and center-crop the image into 64×64×64.
+In MedMNIST+, we center-crop the images of size 120x120x120 into 64×64×64.
 
 ### FractureMNIST3D
 In MedMNIST, for each annotated fracture area, we calculate its center and resize the center-cropped 64×64×64 image into 28×28×28. 
@@ -83,6 +82,6 @@ In MedMNIST, we fix the non-watertight mesh with PyMeshFix and voxelize the wate
 In MedMNIST+, we fix the non-watertight mesh with PyMeshFix and voxelize the watertight mesh with trimesh into 64×64×64 voxels. 
 
 ### SynapseMNIST3D
-In MedMNIST, for each labeled synaptic location, we crop a 3D volume of 1024×1024×1024 and resize it into 28×28×28 voxels. 
+In MedMNIST, the original data is of the size 100um×100um×100um and the resolution 8nm×8nm×30nm, where a 30umx30umx30um sub-volume was used in the MitoEM dataset with dense 3D mitochondria instance segmentation labels. For each labeled synaptic location, we crop a 3D volume of 1024nm×1024nm×1024nm and resize it into 28×28×28 voxels. 
 
 In MedMNIST+, for each labeled synaptic location, we crop a 3D volume of 1024×1024×1024 and center-crop it into 64×64×64 voxels.