MOOSE v3 — Clinical
MOOSE v3 — Clinical (version 3.1.5) is a peer-reviewed CT Image Segmentation model for human medical imaging, available to run on managed, EU-hosted cloud GPUs through Nalvera.AI — upload a scan, run the model, and download reproducible, open-format results with no local GPU or CUDA setup.
Model card
Description
Clinical CT segmentation suite from ENHANCE-PET covering cardiac chambers, digestive tract, lung lobes, muscles, organs, vertebrae, ribs, and body composition.
Training data
MOOSE 3.0 clinical models (ENHANCE-PET / Shiyam Sundar et al.), trained on multi-centre clinical CT (and PET/CT) cohorts with expert-reviewed labels across digestive, lung, rib, peripheral-bone and other tissue groups. Adult human anatomy.
Intended use
Research, tissue-group-specific segmentation of clinical CT for systemic/whole-body analysis and PET/CT quantification workflows. Not a diagnostic device.
Known limitations
Tuned to clinical-CT appearance; degrades on non-clinical protocols, strong artefact, paediatric anatomy, and tissue groups outside the selected task. Group-specific models must be matched to the anatomy actually in the field of view.
Expected runtime
~2–6 min/scan on a modern GPU (medium/large class), task- and coverage-dependent.
Input modality
CT
Species
human
Output format
.nii.gz
Output classes (19)
1 adrenal_gland_left2 adrenal_gland_right3 bladder4 brain5 gallbladder6 kidney_left7 kidney_right8 liver9 lung_lower_lobe_left10 lung_lower_lobe_right11 lung_middle_lobe_right12 lung_upper_lobe_left13 lung_upper_lobe_right14 pancreas15 spleen16 stomach17 thyroid_left18 thyroid_right19 trachea
Citations & references
Research use only — not a medical device. Outputs are not
validated for clinical diagnosis or treatment.
Run this model →