NGC | Catalog
CatalogModelsclara_pt_covid19_ct_lung_annotation

clara_pt_covid19_ct_lung_annotation

For downloads and more information, please view on a desktop device.
Logo for clara_pt_covid19_ct_lung_annotation

Description

A pre-trained model for volumetric (3D) segmentation/annotation of the lung from CT images.

Publisher

NVIDIA

Use Case

Annotation

Framework

Clara Train

Latest Version

1

Modified

August 10, 2021

Size

207.92 MB

Model Overview

A pre-trained model for volumetric (3D) segmentation/annotation of the lung from CT images.

Model Architecture

The model is trained using a 3D anisotropic hybrid network [1].

Diagram showing the flow from model input, through the model architecture, and to model output

Training

The training was performed with the following:

  • Script: train.sh
  • 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

Dataset

This model was trained on a global dataset with a large experimental cohort collected from across the globe. The CT volumes of 120 independent subjects are provided by NIH with experts’ lung region annotations.

  • Target: Lung
  • Task: Segmentation
  • Modality: CT
  • Size: 120 3D volumes (90 Training, 10 Validation, 20 Testing)
  • Challenge: Large ranging foreground size

Performance

Dice score is used for evaluating the performance of the model. On the test set, the trained model achieved score of 0.9414 for lung.

Training

Training loss over 2000 epochs.

Graph that shows training acc over 2000 epochs

Validation

Validation mean dice score over 2000 epochs.

Graph that shows validation mean dice getting higher over 2000 epochs until converging around 0.97

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

Input: 2 channel

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

Preprocessing:

  1. Clip the intensity (CT Hounsfield Unit) between [-1024, 1024] and scale to [0,1]
  2. Crop foreground of image using ground truth label with margin 20
  3. Add channel of Gaussian heatmap using extreme points

Augmentation for training:

  1. Randomly shifting intensity of the volume
  2. Randomly zooming
  3. Resize to 128 x 128 x 128
  4. Randomly spatial flipping
  5. Randomly rotate 90 degrees

Output

Output: 2 channels

  • Label 0: background
  • Label 1: lung

Limitations

This training and inference pipeline was developed by NVIDIA. It is based on a segmentation model developed by NVIDIA researchers in conjunction with the NIH. 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.

References

[1] Liu, Siqi, et al. "3D Anisotropic Hybrid Network: Transferring Convolutional Features from 2D Images to 3D Anisotropic Volumes." In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 851-858). Springer, Cham. https://arxiv.org/pdf/1711.08580.pdf

License

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.