Image Registration

What it does

Image Registration aligns two images: a fixed one and a moving one, from the same or different modalities. It is the step behind every follow-up comparison, from oncology restaging to brain MRI follow-up and prior chest CT, and behind multi-modal fusion such as MR on CT.

The default run computes a rigid or affine transform, the safe choice for most comparisons. Deformable registration is opt-in and returns a displacement field with a folding check. Each run also returns the moving image resampled onto the fixed one, because few PACS display registration objects on their own.

Intended use

Technical support for comparison and fusion: putting a prior study in the same space as the current one, fusing a second modality for review or planning, and propagating regions of interest from one time point to the next. A deformable field can hide real growth if it is used naively to measure change, so rigid or affine alignment is the default and deformable registration is chosen deliberately.

Who it is for

Radiologists and oncologists comparing studies over time, radiotherapy and nuclear-medicine teams fusing modalities, researchers running longitudinal analyses, and developers who need alignment inside their own pipelines.

Accepted input

• Two series: a fixed and a moving image, as DICOM series or NIfTI volumes, of the same or different modality.
• Two studies: a current study in files and a prior in prior_files, for the longitudinal tool below.

Requirements

Requirement	Why
Correct spacing, position and orientation in both images	The transform is computed in patient space
Overlapping anatomy	Registration needs shared structures to match
Similar phase for deformable runs	Different contrast phases or breathing states can be mistaken for deformation

Different fields of view between studies and large anatomical change after surgery or weight loss lower registration quality; the quality checks report it.

Options

options.transform selects rigid, affine or deformable. Without it, the run is rigid or affine.

The result

One result per pair: the transform, the resampled moving image and quality checks.

Section	Content
Rigid and affine registration	Rigid or affine transform and the moving image resampled onto the fixed one
Deformable registration	Displacement field, warped image and a Jacobian-based folding check

Standards

• DICOM: Spatial Registration Storage for rigid and affine transforms; Deformable Spatial Registration Storage for displacement fields; resampled series as new DICOM series.
• Other formats: ITK transform files and NIfTI displacement fields.

Related tools

• utils/register/align: the pairwise run described on this page.
• utils/register/longitudinal: current and prior study in, matched series pairs out, aligned along the body axis by body-part score, registered rigidly and optionally deformably; the transform comes as DICOM REG and propagated masks and regions of interest as SEG.

Quality notes

The folding check reports where the Jacobian determinant of the displacement field is zero or negative, that is, where the deformation folds tissue onto itself; such regions are flagged so that change is not measured there.