The proposed dataset contains clinically acquired multiplanar, multi-sequential MRI slices (MPMSI) which are used as input to the segmentation model without any preprocessing. The proposed AJBDS-2023 consists of 10667 images of real patients imaging data with a size of 320*320*3. Acquired images have T1W, TW2, Flair, T1W contrast, ADC, and DWI sequences. Pixel-based ground-truth annotated images of the tumor core and edema of 6334 slices are made manually under supervision of a radiologist. Quantitative assessment of AJBDS-2023 images is done by a novel U-network on 4333 MRI slices.
Steps to reproduce
This dataset can be used for any research activity making citations 2 "Clinically Acquired New Challenging Dataset for Brain SOL Segmentation: AJBDS-2023 '' with contributing authors.