NGC | Catalog


For downloads and more information, please view on a desktop device.
Logo for clara_pt_spleen_ct_annotation


A pre-trained model for volumetric (3D) annotation of the spleen from CT image.



Use Case



Clara Train

Latest Version



March 25, 2022


36.99 MB

Model Overview

A pre-trained model for volumetric (3D) annotation of the spleen from CT image.

Note: The 4.1 version of this model is only compatible with the 4.1 version of the Clara Train SDK container

Model Architecture

The model is trained to annotate the spleen based on 1 input CT abdominal scan and 6 extreme points. The annotation algorithm is described in [1]. The network architecture is described in [2]. The workflow is described as below.



This model utilized a similar architecture as Residual U-net [2]. The training was performed with the following:

  • Script:
  • GPU: At least 16GB of GPU memory.
  • Actual Model Input: 128 x 128 x 128
  • AMP: True
  • Optimizer: Adam
  • Learning Rate: 2e-4
  • Loss: DiceLoss


The training data is the Task09_Spleen.tar from MSD challenge

  • Target: spleen
  • Task: Annotation
  • Modality: CT
  • Size: 41 3D volumes

The provided labeled data was partitioned, based on our own split, into training (31 studies), validation (4 studies) and testing (6 studies) datasets.


The model was trained with 200 cases with our own split, as shown in the datalist json file in config folder. The achieved Dice scores on the validation and testing data are:

Spleen : 0.954 (testing)





How to Use this Model

The model was validated with NVIDIA hardware and software. For hardware, the model can run on any NVIDIA GPU with memory greater than 16 GB. For software, this model is usable only as part of Transfer Learning & Annotation Tools in Clara Train SDK container. Find out more about Clara Train at the Clara Train Collections on NGC.

Full instructions for the training and validation workflow can be found in our documentation.


Input: 2 channel

  • channel 1: CT
  • channel 2: Gaussian heatmap using extreme points


  1. Clip the intensity (CT Hounsfield Unit) between [-57,164] and scale to [0,1]
  2. Crop foreground of image using ground truth label with margin = 20

Augmentation for training:

  1. Randomly shifting intensity of the volume
  2. Randomly spatial flipping
  3. Randomly rotate 90 degrees
  4. Randomly zoom
  5. Resize to 128 x 128 x 128
  6. Add channel of Gaussian heatmap using extreme points


Output: 1 channel spleen


This training and inference pipeline was developed by NVIDIA. It is based on a segmentation model developed by NVIDIA researchers. This research use only software has not been cleared or approved by FDA or any regulatory agency. Clara pre-trained models are for developmental purposes only and cannot be used directly for clinical procedures.


[1] Maninis, Kevis-Kokitsi, et al. "Deep extreme cut: From extreme points to object segmentation." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.

[2] Kerfoot E., Clough J., Oksuz I., Lee J., King A.P., Schnabel J.A. (2019) Left-Ventricle Quantification Using Residual U-Net. In: Pop M. et al. (eds) Statistical Atlases and Computational Models of the Heart. Atrial Segmentation and LV Quantification Challenges. STACOM 2018. Lecture Notes in Computer Science, vol 11395. Springer, Cham.


End User License Agreement is included with the product. Licenses are also available along with the model application zip file. By pulling and using the Clara Train SDK container and downloading models, you accept the terms and conditions of these licenses.