MedSegDiffV2<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.Medical

MedSegDiff-V2: Spectrum-space diffusion for medical segmentation.

Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → MedicalSegmentationBase<T> → MedSegDiffV2<T>

For Beginners

Medical image segmentation with diffusion models. High-quality boundary delineation.

Common use cases:

  • Medical image segmentation with diffusion models
  • High-quality boundary delineation
  • Uncertain region handling in medical images
  • Multi-scale medical feature extraction

How It Works

Technical Details:

  • Conditional diffusion process for segmentation mask generation
  • Spectrum-Space Transformer (SS-Former) for multi-scale features
  • Anchor condition + step-uncertainty-aware attention
  • Iterative denoising for progressively refined segmentation

Reference: Wu et al., “MedSegDiff-V2: Diffusion-based Medical Image Segmentation with Transformer”, AAAI 2024.

Example

// Create a MedSegDiffV2 model for diffusion-based medical segmentation
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.ThreeDimensional,
    taskType: NeuralNetworkTaskType.BinaryClassification,
    inputHeight: 256, inputWidth: 256, inputDepth: 1, outputSize: 1);
var model = new MedSegDiffV2<double>(architecture);
// Every tunable is now set through the options object; these are the defaults:
var custom = new MedSegDiffV2<double>(architecture,
    options: new MedSegDiffV2Options { NumClasses = 1, DropRate = 0 });
            
// Or load a pre-trained ONNX model for iterative denoising segmentation
var onnxModel = new MedSegDiffV2<double>(architecture, "medsegdiffv2.onnx");

Constructors

Constructor Summary
MedSegDiffV2(NeuralNetworkArchitecture<T> architecture, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction, MedSegDiffV2Options options) Initializes MedSegDiffV2 in native (trainable) mode.
MedSegDiffV2(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, MedSegDiffV2Options options) Initializes MedSegDiffV2 in ONNX (inference-only) mode.

Properties

Property Summary
bool Supports3D { get; } MedSegDiff-V2 operates on 2D slices only, so volumetric input is handled slice-wise.

Methods

Method Summary
MedicalSegmentationResult<T> SegmentFewShot(Tensor<T> queryImage, Tensor<T> supportImages, Tensor<T> supportMasks)
MedicalSegmentationResult<T> SegmentSlice(Tensor<T> slice) Segments a single 2D medical slice.
MedicalSegmentationResult<T> SegmentVolume(Tensor<T> volume) Segments a volume by treating it as a single slice (MedSegDiff-V2 is a 2D model).
ModelMetadata<T> GetModelMetadata() Collects metadata describing this model’s configuration.