Looking for Models for 3D CT Lymph Node Segmentation with Partial/Weak Annotations

Research
1

Hi all,

I’m a master’s student specializing in Big Data with limited clinical knowledge. I have downloaded the TCIA(The Cancer Imaging Archive) Mediastinal-Lymph-Node-SEG dataset, which contains CT scans and partial lymph node segmentations for around 500 cancer patients across various cancer types.

I’m a bit lost on how to perform segmentation on this data and would love your input.

  • I want to build a model that provide clinically useful outputs for radiologists something that can work with partially or weakly annotated 3D CT data and still produce accurate outputs, such as lymph node segmentations or volumetric measurements

  • Any recommendations for agents or pipelines that can help preprocess, organize, and train models efficiently on large .dcm datasets that also perform well in clinical or research settings beyond the usual nnU-Net or U-Net

I’m looking for suggestions on:Latest model architectures or pipelines (segmentation, volumetric quantification, disease progression prediction).

1 Like
2

MONAI is supposedly recommended…?
If your topic involves using ML or AI in science, you might find valuable insights by checking out Hugging Science.