Skip to content

report.md

Latest commit

 

History

History
217 lines (138 loc) · 5.63 KB

report.md

File metadata and controls

217 lines (138 loc) · 5.63 KB

Multimodal Image Registration Report

This report describes a complete multimodal image registration pipeline for four medical images:

  • MRI_T2.dcm
  • MRI_DWI.dcm
  • PET.dcm
  • MRI_T2_rot.dcm

The workflow includes image loading, spatial-resolution harmonization, zero padding, exhaustive translation search, exhaustive rotation search, visual assessment with checkerboard overlays, and evaluation through SSD, NCC, and Mutual Information (MI).

1. Objective

The goal is to align multiple imaging modalities into a common spatial frame so that anatomical structures can be compared consistently across acquisitions.

Main tasks:

  1. Load the DICOM images and inspect size and resolution.
  2. Resample floating images to match reference spatial resolution.
  3. Apply zero padding to obtain common image size.
  4. Estimate best translation for DWI and PET relative to T2.
  5. Estimate best rotation for MRI_T2_rot relative to MRI_T2.
  6. Evaluate results using SSD, NCC, and MI.
  7. Visualize final alignment with checkerboard overlays and summary figures.

2. Reference Image Selection

The T2-weighted MRI image is used as the reference image because:

  • it has the highest in-plane resolution,
  • it provides detailed anatomy for geometric matching.

Input image sizes

Image Size (pixels)
MRI T2 512 x 512
MRI DWI 128 x 128
PET 128 x 128
MRI T2 Rot 512 x 512

Pixel spacing

Image Pixel spacing (mm)
MRI T2 0.449 x 0.449
MRI DWI 1.750 x 1.750
PET 2.5743 x 2.5743
MRI T2 Rot 0.449 x 0.449

3. Resolution Harmonization and Padding

Floating images were resampled to match T2 spacing.

Scaling factors

  • DWI -> T2: 3.8976
  • PET -> T2: 5.7333

Image sizes after scaling

Image Size after scaling
MRI T2 512 x 512
MRI DWI (scaled) 499 x 499
PET (scaled) 734 x 734

After scaling, images were zero-padded to a common field of view.

Rescaling and Padding Overview

4. Translation Registration

Translation registration was performed for:

  • T2 vs DWI
  • T2 vs PET

Search range:

  • dx, dy in [-20, 20]

Metrics computed at each shift:

  • SSD
  • NCC
  • MI

Final translation was selected from NCC optimum.

Best translation-metric values

DWI registration

  • minimum SSD: 1947122952.0000
  • maximum NCC: 0.9329
  • maximum MI: 0.9518

PET registration

  • minimum SSD: 2192681244.0000
  • maximum NCC: 0.9649
  • maximum MI: 0.8851

Selected translations

Floating image Best translation [x y] (pixels)
MRI DWI [-1, -1]
PET [6, -16]

5. Translation Registration Assessment

SSD and NCC heatmaps

SSD and NCC Heatmaps

MI heatmaps

Mutual Information Heatmaps

Checkerboard overlays

Checkerboard T2 vs DWI

Checkerboard T2 vs PET

6. Rotation Registration

Rotation registration was performed between:

  • MRI_T2.dcm
  • MRI_T2_rot.dcm

Angular search:

  • range: 0 deg to 359.5 deg
  • step: 0.5 deg

All three metrics identified the same optimal angle:

  • SSD minimum at 340.00 deg
  • NCC maximum at 340.00 deg
  • MI maximum at 340.00 deg

Rotation comparison

Rotation Registration T2 vs T2rot

7. Rotation Registration Assessment

Joint histogram after registration

Joint Histogram T2 vs T2rot Registered

Checkerboard overlay after rotation correction

Checkerboard T2 vs T2rot Registered

Metric trends over angle

Rotation Metric Trends

8. Final Mutual Information Values

Registered pair Final MI
T2 vs T2_rot_registered 2.9144
T2 vs DWI_registered 0.9471
T2 vs PET_registered 0.8837

Final Mutual Information Bar Chart

9. Pairwise Registration Reports

T2 vs DWI

Registration Report T2-DWI

T2 vs PET

Registration Report T2-PET

T2 vs T2rot

Registration Report T2-T2rot

10. Final Registration Summary Table

Image Scaling factor Translation [x y] (pixels) Rotation
MR_T2.dcm REF [0, 0] -
MRI_DWI.dcm 3.8976 [-1, -1] -
PET.dcm 5.7333 [6, -16] -
MRI_T2_rot.dcm - - 340.00 deg

11. Files Included

  • Medical_Image_Registration.m - full MATLAB script for the registration workflow.
  • lib/metrics/SSD.m - Sum of Squared Differences.
  • lib/metrics/NCC.m - Normalized Cross Correlation.
  • lib/metrics/MI.m - Mutual Information.
  • lib/operators/zeroPadding.m - field-of-view harmonization helper.
  • lib/visualization/checkerboard_view.m - visual comparison helper.
  • figures/ - full set of exported registration figures.

12. Conclusion

The workflow successfully performs multimodal registration between T2, DWI, PET, and rotated T2 images.

Key outcomes:

  • T2 was used as the anatomical reference.
  • DWI and PET were resampled and padded before translation search.
  • Best translations were [-1, -1] for DWI and [6, -16] for PET.
  • Best rotation for MRI_T2_rot was 340.00 deg.
  • Heatmaps, checkerboards, and summary reports confirmed good final alignment.