Brain Fat Suppression
AI-powered fat signal suppression for brain MRI. Our 3D U-Net fuses multiple MRI sequences — T1, T2 and FLAIR — to computationally suppress fat artefacts, delivering cleaner diagnostic images without requiring specialised pulse sequences.
Why Brain Fat Suppression?
Fat signals in brain MRI obscure critical tissue boundaries and mimic pathology. Traditional hardware-based suppression is time-consuming, prone to artefacts, and not always available. BrainFS solves this computationally.
Clearer Tissue Contrast
Removes fat signal contamination to reveal true tissue boundaries, improving visibility of lesions, oedema, and subtle pathological changes.
Multi-Modal Fusion
Combines information from T1, T2 and FLAIR sequences simultaneously, leveraging complementary contrast properties for superior suppression quality.
True 3D Consistency
Processes entire volumes natively in 3D, maintaining spatial coherence across slices — eliminating the inter-slice artefacts of 2D approaches.
No Extra Scan Time
Works with standard sequences already acquired in routine protocols. No additional pulse sequences, no extended scan sessions, no patient discomfort.
Artefact Reduction
Attention-driven processing selectively targets fat signals while preserving genuine anatomical detail, avoiding the chemical-shift artefacts of hardware methods.
Format Flexibility
Accepts both DICOM and NIfTI inputs, integrating seamlessly with hospital PACS systems and research workflows alike.
How It Works
BrainFS uses a 3D U-Net with CBAM attention modules to learn the mapping from multi-sequence input volumes to fat-suppressed output, trained with perceptual and pixel-level losses.
Multi-Sequence Input
T1, T2 and FLAIR volumes are loaded from DICOM or NIfTI format.
Register & Normalise
Volumes are aligned, resampled and normalised for consistent 3D processing.
3D U-Net + Attention
A 4-level encoder-decoder with CBAM attention at each level separates fat from tissue.
Fat Signal Removal
Perceptual and pixel-level losses ensure fat is suppressed while genuine anatomy is preserved.
Clean Output
A fat-suppressed 3D volume with preserved tissue contrast, ready for diagnosis.
Clinical Applications
BrainFS helps radiologists across diagnostic scenarios where fat signals compromise image quality.
Tumour Margin Delineation
Clear fat signals from peri-tumoural regions to better visualise true lesion boundaries, improving surgical planning and treatment response monitoring.
Orbit & Skull Base Imaging
Fat-rich regions around the orbits and skull base routinely obscure pathology. BrainFS computationally suppresses these signals without additional scan time.
Retrospective Enhancement
Apply fat suppression to historical scans where the original protocol did not include it — enabling longitudinal comparison with modern fat-suppressed studies.