Abdominal MRI

What it does

Abdominal MRI reads the ordered "MRI abdomen" exam in its liver, pancreas, MRCP and kidney protocols. One model covers them all: the sequences in the study decide which sections run. Every study gets organ masks and volumes; a dynamic liver study adds lesion detection and classification; an MRCP adds the ducts; chemical-shift or PDFF/R2* maps add fat and iron quantification.

The liver read covers diffuse disease (PDFF, R2*, estimated iron concentration and steatosis grade), focal lesions classified as cyst, haemangioma, FNH, HCC, cholangiocarcinoma or metastasis, LI-RADS categories for patients at risk, and Couinaud-segment volumes with the future liver remnant before resection. Outside the liver, the run grades pancreatic cysts against the Fukuoka 2017 and ACR rules, classifies cystic renal masses with Bosniak v2019 and measures total kidney volume in ADPKD.

Intended use

Decision support for physicians reporting abdominal MRI, where LI-RADS and cyst-management rules are often misapplied: per-observation LI-RADS features, follow-up intervals tied to the guideline that sets them, and quantification that is impractical by hand. Every lesion and category comes with its evidence.

Who it is for

Abdominal and general radiologists, hepatologists running HCC surveillance and MASLD clinics, hepatobiliary surgeons planning resections, gastroenterologists and nephrologists following ADPKD. Developers use organ volumes and PDFF for trials and population studies.

Accepted input

• DICOM MR study (MR Image, Enhanced MR or Legacy Converted Enhanced MR) as a zip or by DICOMweb STOW-RS; NIfTI with a BIDS-style sidecar on the API.
• Vendor PDFF and R2 maps* or multi-echo chemical-shift (Dixon) source images.
• One prior in prior_files for LI-RADS threshold growth and cyst follow-up.

Sequences per section

Section	Needs
Organ segmentation	Any abdominal sequence
Liver fat and iron	PDFF/R2* maps or multi-echo chemical shift
Liver lesions and LI-RADS	Multiphase T1 (pre, arterial, portal venous, delayed or hepatobiliary), T2, DWI
Liver volumetry	T1 or another volumetric sequence
MRCP	3D MRCP, ± thick-slab 2D
Pancreatic cysts	T2 and post-contrast, ideally with MRCP
Renal cysts (Bosniak)	T2 and pre- and post-contrast T1
Total kidney volume	Coronal or axial T2

The contrast agent comes from ContrastBolusAgent: a hepatobiliary agent and an extracellular agent change how washout and the late phases are judged, and late-arterial timing is checked on the dynamic series.

Optional context

Cirrhosis or other LI-RADS risk factors, height (height-adjusted TKV), the intended resection extent and the question go in clinical_context. LI-RADS categories are only assigned when the patient is at risk.

The result

Section	Content
Organ segmentation	40–50 organ and structure masks with volumes
Liver fat and iron	Whole-liver and per-segment PDFF and R2*, estimated liver iron concentration, steatosis grade
Liver lesions	Detection, segmentation and class of each focal lesion
LI-RADS	Per observation: size, APHE, washout, capsule, threshold growth and LR category including LR-M and LR-TIV
Liver volumetry	Total and Couinaud-segment volumes, future liver remnant %
MRCP	Biliary and pancreatic duct segmentation, calibre profile, strictures, dilatation and filling defects
Pancreatic cysts	Cyst detection, size, duct communication, worrisome features and high-risk stigmata, follow-up interval
Renal cysts (Bosniak)	Cystic renal mass features and Bosniak class
Total kidney volume	TKV, height-adjusted TKV and Mayo imaging class
Draft report	Opt-in (options.report): an English narrative that interprets the result for the reader (findings, impression, limitations); every number is checked against the findings

Standards

• DICOM: SEG for organs, segments and lesions; SR TID 1500 per observation with tracking identifiers; PDFF and R2* as Parametric Maps; KOS key images and Encapsulated PDF. Liver results carry a couinaud block.
• Liver: LI-RADS v2018 and the QIBA PDFF profile.
• Pancreas: Fukuoka 2017 consensus and the ACR incidental pancreatic cyst recommendations.
• Kidney: Bosniak v2019 and the Mayo ADPKD imaging classification.

Quantitative maps stay in physical units and are never bias-corrected.