This repository holds the code for the MIDL 2024 short paper submission "Automatic classification of prostate MR series type using image content and metadata".
With the wealth of medical image data, efficient curation is essential. Assigning the sequence type to magnetic resonance images is necessary for scientific studies and artificial intelligence-based tasks. Incomplete or missing metadata prevents effective automation, leading to time-consuming and error-prone processes by clinicians. We propose a deep-learning method for classification of prostate cancer scanning sequences based on a combination of image data and DICOM metadata. We demonstrate superior results compared to metadata or image data alone, and make our code publicly available at https://github.com/deepakri201/DICOMScanClassification.
DICOMTagClassification_ImageAndMetadata_setup_exp3.ipynb This notebook is for querying Imaging Data Commons data, and obtaining the appropriate metadata from the DICOM files.
DICOMTagClassification_ImageAndMetadata_train_and_test_exp3.ipynb This notebook is for the training and test data setup, running training/validation for the three models, and for running inference. Figures and tables are also produced in this notebook.
Further details will be added to this page about how to use the code.
Deepa Krishnaswamy ([email protected]), April 2024, Brigham and Women's Hospital