update medmnist+ details

README.md

@@ -175,8 +175,8 @@ Please also cite source data paper(s) of the MedMNIST subset(s) as per the descr
 # Release Notes
 * [WIP] `v3.0.0`: MedMNIST+ featuring larger sizes: 64x64, 128x128, and 224x224 for 2D, and 64x64x64 for 3D.
     - [ ] TODO: test all (`__main__.py`, `getting_started.ipynb`, `getting_started_without_PyTorch`, including using `Evaluator`)
-    - [ ] TODO: new examples for v3 large size (in `getting_started.ipynb`, `getting_started_without_PyTorch`, including using `Evaluator`)
-    - [ ] TODO: a [document](on_medmnist_plus.md) on MedMNIST+ details
+    - [x] TODO: new examples for v3 large size (in `getting_started.ipynb`, `getting_started_without_PyTorch`, including using `Evaluator`)
+    - [x] TODO: a [document](on_medmnist_plus.md) on MedMNIST+ details
     - [ ] TODO: clean README, release on PyPI.
 * `v2.2.4`: removed a small number of blank samples in OrganAMNIST, OrganCMNIST, OrganSMNIST, OrganMNIST3D, and VesselMNIST3D. 
 * `v2.2.3`: DermaMNIST license to CC BY-NC 4.0

on_medmnist_plus.md

@@ -1,3 +1,88 @@
 # How we generated the MedMNIST+ (the larger-size version of MedMNIST)
 
-#TODO 
+The data in MedMNIST+ corresponds one-to-one with that in MedMNIST (size 28x28 for 2D, 28x28x28 for 3D), with the only difference lying in the resolution of the images. In other words, the dataset splits (*i.e.* train, val, and test) and the sample indices are the same for MedMNIST and MedMNIST+. 
+
+It is important to emphasize that for the data obtained through center-crop followed by resizing in MedMNIST, we use the same size for center-crop and subsequently resize to obtain the target resolution in MedMNIST+. 
+
+We will describe the details of MedMNIST+, compared with MedMNIST. More details of MedMNIST can be found in our [paper](https://doi.org/10.1038/s41597-022-01721-8).
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+### 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. 
+
+In MedMNIST+, following the same procedure in MedMNIST, we obtain 2D maximum projections by taking the maximum pixel value along the axial-axis of each pixel, resulting in images of size 1x32x32. Then, for size 224x224, 128x128, and 64x64, we resize the source images into the target size. 
+
+### Organ{A,C,S}MNIST
+In MedMNIST, Hounsfeld-Unit (HU) of the 3D images are transformed into gray-scale with an abdominal window. We crop 2D images from the center slices of the 3D bounding boxes in axial/coronal/sagittal views (planes). The only diferences of Organ{A,C,S}MNIST are the views. The images are resized into 1×28×28 to perform multi-class classifcation of 11 body organs. 
+
+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+, we directly use the 3D bounding boxes 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+, 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.
+
+### 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 calculate the center of adrenal and center-crop the image 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. 
+
+In MedMNIST+, for each annotated fracture area, we calculate its center and center-crop the image into 64×64×64. 
+
+### VesselMNIST3D
+In MedMNIST, we fix the non-watertight mesh with PyMeshFix and voxelize the watertight mesh with trimesh into 28×28×28 voxels. 
+
+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+, for each labeled synaptic location, we crop a 3D volume of 1024×1024×1024 and center-crop it into 64×64×64 voxels.