Skip to content

Data_Loading_and_Preprocessing_Module.md

Latest commit

 

History

History
106 lines (60 loc) · 5.05 KB

Data_Loading_and_Preprocessing_Module.md

File metadata and controls

106 lines (60 loc) · 5.05 KB
graph LR
    BrainS18Dataset["BrainS18Dataset"]
    __getitem__["__getitem__"]
    __training_data_process__["__training_data_process__"]
    __testing_data_process__["__testing_data_process__"]
    __nii2tensorarray__["__nii2tensorarray__"]
    __itensity_normalize_one_volume__["__itensity_normalize_one_volume__"]
    __resize_data__["__resize_data__"]
    __crop_data__["__crop_data__"]
    BrainS18Dataset -- "manages" --> __getitem__
    BrainS18Dataset -- "provides access" --> __getitem__
    __getitem__ -- "orchestrates" --> __training_data_process__
    __getitem__ -- "invokes" --> __training_data_process__
    __getitem__ -- "orchestrates" --> __testing_data_process__
    __getitem__ -- "invokes" --> __testing_data_process__
    __getitem__ -- "uses" --> __nii2tensorarray__
    __training_data_process__ -- "utilizes" --> __itensity_normalize_one_volume__
    __training_data_process__ -- "utilizes" --> __resize_data__
    __training_data_process__ -- "utilizes" --> __crop_data__
    __testing_data_process__ -- "utilizes" --> __itensity_normalize_one_volume__
    __testing_data_process__ -- "utilizes" --> __resize_data__
    __testing_data_process__ -- "utilizes" --> __crop_data__
Loading

CodeBoardingDemoContact

Details

Analysis of the Data Loading and Preprocessing Module in MedicalNet project.

BrainS18Dataset

The foundational component, acting as the primary interface for accessing and managing 3D medical image data. It initializes the dataset, handles file path management, and provides the entry point for data retrieval.

Related Classes/Methods:

getitem

Orchestrates the retrieval and preprocessing of a single data sample. It dynamically selects and applies either the training or testing preprocessing pipeline based on the dataset's current mode.

Related Classes/Methods:

training_data_process

Encapsulates the complete preprocessing pipeline specifically for training data. This includes steps like data cleaning, intensity normalization, resizing, and augmentation (e.g., random cropping).

Related Classes/Methods:

testing_data_process

Encapsulates the standardized preprocessing pipeline for testing or validation data. This pipeline typically includes normalization, resizing, and cropping, but generally excludes aggressive augmentation to ensure consistent evaluation.

Related Classes/Methods:

nii2tensorarray

Converts processed NumPy array data (derived from NIfTI files) into the PyTorch tensor format, making it compatible for direct use with PyTorch models.

Related Classes/Methods:

itensity_normalize_one_volume

Standardizes the intensity values of a single 3D image volume to a predefined range, crucial for consistent model input and performance.

Related Classes/Methods:

resize_data

Adjusts the spatial dimensions of the 3D medical image data to a uniform size, ensuring consistency across the dataset.

Related Classes/Methods:

crop_data

Extracts a specific sub-region from the image data. This can be used for focusing on regions of interest or as a data augmentation technique.

Related Classes/Methods:

FAQ