ct fm feature extractor EU AI Act Compliance Profile

CT-FM Feature Extractor

This model is a feature extractor for CT-FM, a model pre-trained using contrastive self-supervised learning on a huge dataset of 148,000 CT scans from the Imaging Data Commons.

The backbone is based on a SegResNet, a 3D U-Net variant. If you want to just load the model and fine-tune, ignore the feature extraction workflow.

Running instructions

CT-FM Feature Extractor

This notebook demonstrates how to:

1. Load a SSL pre-trained model
2. Set up preprocessing and postprocessing pipelines
3. Perform inference on CT volumes
4. Plot distribution of features extracted

Setup

Install requirements and import necessary packages

# Install lighter_zoo package
%pip install lighter_zoo -U -qq

# Imports
import torch
from lighter_zoo import SegResEncoder
from monai.transforms import (
    Compose, LoadImage, EnsureType, Orientation,
    ScaleIntensityRange, CropForeground
)
from monai.inferers import SlidingWindowInferer

Load Model

Download and initialize the pre-trained model from HuggingFace Hub

# Load pre-trained model
model = SegResEncoder.from_pretrained(
    "project-lighter/ct_fm_feature_extractor"
)
model.eval()

Setup Processing Pipelines

Define preprocessing transforms

# Preprocessing pipeline
preprocess = Compose([
    LoadImage(ensure_channel_first=True),  # Load image and ensure channel dimension
    EnsureType(),                         # Ensure correct data type
    Orientation(axcodes="SPL"),           # Standardize orientation
    # Scale intensity to [0,1] range, clipping outliers
    ScaleIntensityRange(
        a_min=-1024,    # Min HU value
        a_max=2048,     # Max HU value
        b_min=0,        # Target min
        b_max=1,        # Target max
        clip=True       # Clip values outside range
    ),
    CropForeground()    # Remove background to reduce computation
])

Your obligations as a ct fm feature extractor deployer

Track your progress

Sign up to track obligations, receive deadline reminders, and generate audit packages.

Tags

Languages

Library