MedicalSegmentationOutput<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.Common

Output for medical image segmentation with volumetric support and clinical metadata.

For Beginners

Medical segmentation output includes organ/structure labels for each pixel or voxel, along with clinical metrics like Dice scores and volume measurements that doctors use to assess segmentation quality and make clinical decisions.

Constructors

Constructor Summary
MedicalSegmentationOutput()

Properties

Property Summary
Dictionary<int, double> DiceScores { get; set; } Per-class Dice scores (if ground truth was provided during evaluation).
Dictionary<int, double> HausdorffDistances { get; set; } Per-class Hausdorff distances in mm (if ground truth was provided).
Dictionary<int, double> SurfaceDiceScores { get; set; } Per-class surface Dice scores at specified tolerance (if ground truth was provided).
List<AnatomicalStructure> Structures { get; set; } Segmented anatomical structures with volume and metadata.
Tensor<T> Labels { get; set; } Per-pixel/voxel class labels.
Tensor<T> Probabilities { get; set; } Per-class probability maps.
Tensor<T> UncertaintyMap { get; set; } Uncertainty map [H, W] or [D, H, W] indicating model uncertainty per pixel/voxel.
TimeSpan InferenceTime { get; set; } Inference time.
bool Is3D { get; set; } Whether this result is from a 3D volumetric segmentation.
double OverallConfidence { get; set; } Model confidence summary across all structures.
double[] OrientationMatrix { get; set; } Image orientation matrix (for DICOM compatibility).
double[] VoxelSpacing { get; set; } Voxel spacing in mm [x, y, z] (for 3D results).
string Modality { get; set; } Imaging modality used (CT, MRI_T1, MRI_T2, etc.).