ANTsPy MedAI

ANTsPy Longitudinal MRI Registration and Subtraction

Automated co-registration and quantitative assessment of brain metastases

Preloaded Pairs:

Resolution:

Transform:

Preloaded MRI Scans Repository (Click to select for Pre or Post) Hide Scans

PRE

Scan Label: -

Sequence: -

Dimensions: -

POST

Scan Label: -

Sequence: -

Dimensions: -

DIFF Difference and Quantification

Tumor Regression (Signal Loss)

Post-op Cavity / Modification

Estimated Lesion Regression

- %

Spatial Cross-Correlation

- %

Radiological Evaluation (RANO Criteria):

Computing registration...

Methodology and Algorithmic Pipeline

How ANTsPy co-registration, subtraction, and RANO response assessment work

1. Clinical Motivation

Serial follow-up MRI scans of melanoma brain metastases (MaMelanom MTS) frequently suffer from head tilt, translation, and rotational misalignment across acquisitions. This platform automates spatial alignment to baseline coordinates and isolates true pathological signal evolution from positional artifacts.

2. Four-Step Algorithmic Pipeline

Step 1: Spatial Co-Registration (ANTsPy)

Advanced Normalization Tools (ANTsPy) executes Rigid (translation + rotation) or Affine registration using Mattes Mutual Information metric. The follow-up scan (Post) is transformed (warped) into the exact anatomical reference coordinates of the baseline (Pre) scan.

Step 2: Intensity Normalization and Masking

A robust Otsu skull/brain mask isolates intracranial tissue from air and background noise. Non-linear contrast differences and MR gain variations are normalized using robust percentile and z-score intensity calibration.

Step 3: Differential Signal Subtraction (Diff Heatmap)

Pixel-wise signal subtraction is computed: Diff = Post(warped) - Pre.
• Green overlay: Signal Regression / Lesion shrinkage or necrotic resolution.
• Red overlay: Signal Progression / New enhancing lesion or recurrence.

Step 4: Quantitative Metrics and RANO Response

Automated thresholding computes the percentage area change of the target enhancing lesion. Clinical classification follows standard RANO-BM (Response Assessment in Neuro-Oncology Brain Metastases) criteria:

CR Complete Response: 100% disappearance of enhancing lesion.

PR Partial Response: Greater than or equal to 50% decrease in lesion area.

SD Stable Disease: Less than 50% decrease to less than 25% increase.

PD Progressive Disease: Greater than or equal to 25% increase or new lesions.

3. Interactive Controls

• Preloaded Pairs: 5 one-click clinical pairs representing different lesion sizes and therapeutic responses.
• View Toggles: Toggle between ANTsPy Warped vs Raw Unaligned in the Post column, and Subtraction Heatmap vs Fusion Overlay in the Difference column.
• Custom Upload: Test any custom baseline and follow-up MRI slice pair directly.

Authors and Clinical Inquiries

Methodology and Algorithmic Pipeline

Upload Custom Scans for Registration