The present dataset offers a unique corpus of 2D sagittal-view RT-MRI videos along with synchronized audio for 75 subjects performing linguistically motivated speech tasks, alongside the corresponding first-ever public domain raw RT-MRI data.
The dataset also includes 3D volumetric vocal tract MRI during sustained speech sounds and high-resolution static anatomical T2-weighted upper airway MRI for each subject. Funding National Science Foundation Grant Usage metrics. Categories Medical Devices Linguistic Processes incl.
Licence CC BY 4. Hide footer. CT scans store raw voxel intensity in Hounsfield units HU. They range from to above in this dataset. Above are bones with different radiointensity, so this is used as a higher bound. A threshold between and is commonly used to normalize CT scans.
Here we define several helper functions to process the data. These functions will be used when building training and validation datasets. Read the scans from the class directories and assign labels.
Downsample the scans to have shape of xx Rescale the raw HU values to the range 0 to 1. Lastly, split the dataset into train and validation subsets.
The CT scans also augmented by rotating at random angles during training. Since the data is stored in rank-3 tensors of shape samples, height, width, depth , we add a dimension of size 1 at axis 4 to be able to perform 3D convolutions on the data. The new shape is thus samples, height, width, depth, 1. There are different kinds of preprocessing and augmentation techniques out there, this example shows a few simple ones to get started.
While defining the train and validation data loader, the training data is passed through and augmentation function which randomly rotates volume at different angles. Note that both training and validation data are already rescaled to have values between 0 and 1. To make the model easier to understand, we structure it into blocks.
The architecture of the 3D CNN used in this example is based on this paper. It is important to note that the number of samples is very small only and we don't specify a random seed.
As such, you can expect significant variance in the results.
0コメント