DICOM reading, measurement & annotation
Window, scroll, measure and annotate studies to reference findings.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- MicroDicom · Horos · Weasis · RadiAnt
- Class GUI-need label
G3· GUI-only
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R09.01.01.01-T01 Measure RECIST target lesions at baseline and follow-up in Horos
Original instruction
In Horos open de-identified CT studies CT_baseline and CT_followup (patient ONC-017, synthetic case). Measure the longest axial diameter of the 3 target lesions marked on the baseline key images (liver, lung, node short axis for the node), find and measure the same lesions on follow-up, and write output/recist.json with sums, percent change and response category.
- Platforms
mac· macOS- Applications
- Horos
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometrynav· Pan/zoom/scrollvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Checker reads Horos ROI exports and recist.json: each measurement within 2 mm of radiologist gold on the correct slice (+-1); sums and percent change consistent; category (PR) exact. The VM has no shell, pydicom or DICOM toolkits, and reads of the DICOM folder by any process other than Horos are caught by file-access auditing.
Proposed partial credit
35% baseline measurements; 35% follow-up lesions matched and measured; 15% sums/percent; 15% category
R09.01.01.01-T02 Find the largest lung nodule with a lung window in Weasis
Original instruction
Open chest CT study CHEST-044 (synthetic, de-identified) in Weasis. Apply window width 1500 / level -600, scroll the axial series, find the largest solid pulmonary nodule, measure its longest diameter, and save the measurement as a DICOM SR/presentation state; write output/nodule.json with series, slice index and diameter in mm.
- Platforms
win· Windowsmac· macOS- Applications
- Weasis
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrollgeom· Precise geometryvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Saved presentation state shows W/L 1500/-600; slice index within +-1 of gold; diameter within 1 mm; nodule location (measurement midpoint) inside the gold nodule mask. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so images are reachable only in Weasis; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
15% window setting; 40% correct nodule and slice; 30% diameter; 15% saved annotation
R09.01.01.01-T03 Measure liver and spleen attenuation with circular ROIs in RadiAnt
Original instruction
In RadiAnt DICOM Viewer open unenhanced abdominal CT ABD-102 (synthetic). Place 1 cm^2 circular ROIs in the right liver lobe (avoiding vessels) and the spleen on the same slice level, and write output/hu.json with mean HU for each and the liver-minus-spleen difference.
- Platforms
win· Windowsmac· macOS- Applications
- RadiAnt
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometryvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
RadiAnt annotations exported: ROI areas 0.8-1.2 cm^2 inside the gold organ masks and not overlapping vessel masks; means within 5 HU of the checker's recomputation for those ROIs and within 8 HU of gold organ means. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so pixel values are reachable only in RadiAnt; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
30% ROI placement valid; 40% HU values; 30% difference reported
R09.01.01.01-T04 Export anonymized series as full-resolution images in MicroDicom
Original instruction
In MicroDicom open study MSK-009 and export series 2-4 to output/export/ as anonymized DICOM (PatientName, PatientID, BirthDate, InstitutionName removed) and series 3 additionally as full-resolution PNG without burned-in overlay text.
- Platforms
win· Windows- Applications
- MicroDicom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP09· Files & folders- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Checker parses exported DICOM: the 4 tags absent/empty, pixel data identical to source, file counts per series exact; PNGs match native matrix size with no overlay text (OCR finds none). The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so anonymized export must be done in MicroDicom; study-folder access by other processes is caught by file-access auditing.
Proposed partial credit
40% anonymized tags; 20% counts and pixel data; 40% PNG export correct
R09.01.01.01-T05 Box pneumothoraces on six chest radiographs in MicroDicom
Original instruction
Six synthetic-PHI chest radiographs CXR-01..06 are loaded in MicroDicom. For each, draw a rectangle annotation around any pneumothorax and save; write output/boxes.tsv with image, x1, y1, x2, y2 in pixel coordinates, or 'none' when absent.
- Platforms
win· Windows- Applications
- MicroDicom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometryvis· Visual judgmentnav· Pan/zoom/scroll- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Per image: presence classification correct; for positives the box IoU>=0.5 with gold; TSV coordinates must match MicroDicom's saved annotations (within 2 px). The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so images are reachable only in MicroDicom; reads of the image folder by other processes are caught by file-access auditing.
Proposed partial credit
40% presence correct for 6 images; 45% box IoU on positives; 15% TSV consistent with saved annotations
R09.01.01.01-T06 Measure the aortic root on an oblique MPR in Horos
Original instruction
In Horos open contrast CT angiogram CTA-AO-21 (synthetic). Use 3D MPR to align an oblique plane perpendicular to the aortic root at the sinuses of Valsalva and measure the maximal sinus-to-sinus diameter and the sinotubular junction diameter. Save the MPR measurements and write output/aorta.json.
- Platforms
mac· macOS- Applications
- Horos
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometrydrag· Drag & dropvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 5 / 5
Proposed verifier
Exported MPR plane normal within 10 degrees of gold root axis; diameters within 2 mm of gold; measurement positions on the correct anatomical level (z within 5 mm). The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so the volume is reachable only in Horos; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
35% plane orientation; 40% sinus diameter; 25% STJ diameter
R09.01.01.01-T07 Count new FLAIR lesions on synchronized prior and current MRI in Weasis
Original instruction
Open brain MRI studies MS-033 prior and current (synthetic MS cases) side by side in Weasis with synchronized scrolling. Identify FLAIR lesions present on current but not prior, place an arrow annotation on each new lesion, and write output/new_lesions.json with count and slice numbers.
- Platforms
win· Windowsmac· macOS- Applications
- Weasis
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrollvis· Visual judgmentclick· Point & click- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Count equals gold (4); each arrow tip lies within 5 mm of a gold new-lesion centroid on the right slice (+-1); detection F1 computed from arrows. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so images are reachable only in Weasis; reads of the study folders by other processes are caught by file-access auditing.
Proposed partial credit
30% correct count; 55% arrow detection F1; 15% synchronized layout saved
R09.01.01.01-T08 Measure a scoliosis Cobb angle in RadiAnt
Original instruction
In RadiAnt open standing spine radiograph SCOL-12 (synthetic). Identify the end vertebrae of the major thoracic curve and measure the Cobb angle with the Cobb angle tool. Write output/cobb.json with upper and lower end vertebra labels and the angle.
- Platforms
win· Windowsmac· macOS- Applications
- RadiAnt
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometryvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
End vertebrae exact (or within one level, half credit); angle within 3 degrees of gold consensus; RadiAnt annotation export contains a Cobb measurement whose lines lie on the reported vertebrae. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so the image is reachable only in RadiAnt; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
40% end vertebrae; 50% angle; 10% annotation consistent
R09.01.01.01-T09 Estimate a renal cyst volume with slice-by-slice ROIs in Horos
Original instruction
In Horos open CT KID-58 (synthetic). Draw closed polygon ROIs outlining the left renal cyst on every axial slice where it appears, then compute the volume with Horos' ROI volume tool and write output/cyst.json with volume in mL and slice range.
- Platforms
mac· macOS- Applications
- Horos
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
path· Continuous pathnav· Pan/zoom/scrollvis· Visual judgment- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Exported ROI series rasterized: Dice>=0.85 vs gold cyst mask; slice range within +-1; reported volume within 10% of gold. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so the volume is reachable only in Horos; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
40% Dice; 20% slice range; 40% volume
R09.01.01.01-T10 Mark key images with arrow annotations and save as KO and GSPS in Weasis
Original instruction
In Weasis review abdominal CT ABD-311 (synthetic). Mark the 4 key images showing appendicitis findings (dilated appendix, periappendiceal fat stranding, appendicolith, free fluid) with an arrow and short text label each, and save them as a DICOM Key Object Selection plus presentation states.
- Platforms
win· Windowsmac· macOS- Applications
- Weasis
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgmenttype· Text & forms- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Checker parses KO (4 referenced instances, slices within +-2 of gold) and GSPS graphics: arrow tips within 10 mm of gold findings, labels match the 4 finding names. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so KO/GSPS objects can only be created in Weasis; study-folder access by other processes is caught by file-access auditing.
Proposed partial credit
30% KO with 4 images; 50% arrows at findings; 20% labels
R09.01.01.01-T11 Read acquisition parameters from DICOM headers in MicroDicom
Original instruction
Open study CT-MULTI-07 in MicroDicom, which has 5 series. Using the DICOM tag viewer, fill output/params.csv with series number, description, slice thickness, kVp, reconstruction kernel and contrast phase (from SeriesDescription or ContrastBolusAgent).
- Platforms
win· Windows- Applications
- MicroDicom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP02· Records/table- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
CSV values equal header values for all 25 cells (contrast phase normalized to arterial/venous/delayed/none). DICOM libraries are not installed in the sandbox; direct file reads outside MicroDicom are logged.
Proposed partial credit
40% series 1-2 rows; 40% series 3-4 rows; 20% series 5 row (equal share per cell)
R09.01.01.01-T12 Find the SUVmax of the hottest lesion on fused PET/CT in RadiAnt
Original instruction
In RadiAnt open PET/CT study LYM-005 (synthetic). Display PET fused over CT, find the most FDG-avid lesion outside physiologic uptake (brain, bladder, myocardium), place a spherical/circular ROI and report SUVmax (body weight) and lesion location in output/pet.json.
- Platforms
win· Windowsmac· macOS- Applications
- RadiAnt
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrollvis· Visual judgmentgeom· Precise geometry- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Location must be inside the gold hottest-lesion mask; SUVmax within 10% of gold; physiologic sites excluded (choosing bladder scores 0). The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so PET values are reachable only in RadiAnt; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
15% fusion displayed; 50% correct lesion; 35% SUVmax
R09.01.01.01-T13 Create a bone-only volume rendering at a set view in Horos
Original instruction
In Horos open pelvis CT PEL-14 (synthetic), create a 3D volume rendering with a bone preset (soft tissue removed), orient it to an anterior view rotated 30 degrees to the left, and export a 1024x1024 image to output/vr.png; save the 3D view state.
- Platforms
mac· macOS- Applications
- Horos
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropgeom· Precise geometryvis· Visual judgment- Work product
WP03· Visual layout- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Exported PNG 1024x1024; SSIM>=0.80 vs reference render with the same preset and view; saved view camera azimuth within 5 degrees of target; no soft tissue (mean intensity in a soft-tissue mask region below threshold). The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so the volume is reachable only in Horos; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
20% export size; 30% bone-only preset; 25% view angle; 25% SSIM
R09.01.01.01-T14 Grade internal carotid stenosis by NASCET on CT angiography in Weasis
Original instruction
In Weasis open neck CTA CAR-09 (synthetic). Using MPR, measure the narrowest residual lumen diameter of the left ICA and the normal distal ICA diameter, compute percent stenosis by NASCET, and write output/stenosis.json with both diameters and the percent.
- Platforms
win· Windowsmac· macOS- Applications
- Weasis
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometrynav· Pan/zoom/scrollvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 5 / 5
Proposed verifier
Diameters within 0.5 mm of gold measured perpendicular to the vessel axis; stenosis within 5 percentage points; measurements on the correct vessel (left ICA) checked by location. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so the CTA is reachable only in Weasis; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
20% correct vessel and location; 30% residual lumen; 25% distal diameter; 25% percent stenosis
R09.01.01.01-T15 Locate five anatomical landmark slices in a 300-slice CT in MicroDicom
Original instruction
Open chest-abdomen CT CAP-200 (synthetic, 300 axial slices) in MicroDicom. Find the first slice (cranial to caudal) where each of these appears: aortic arch top, carina, left renal hilum, aortic bifurcation, femoral heads. Write output/landmarks.json with slice numbers.
- Platforms
win· Windows- Applications
- MicroDicom
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrollvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Each slice within +-2 of radiologist gold; slice numbering by InstanceNumber. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so slices are reachable only in MicroDicom; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
20% aortic arch top; 20% carina; 20% left renal hilum; 40% aortic bifurcation and femoral heads (equal halves)
R09.01.01.01-T16 Measure fetal biparietal diameter on the right cine frame in RadiAnt
Original instruction
In RadiAnt open ultrasound cine clip OB-22 (synthetic). Scroll to the frame showing the standard transthalamic plane, measure the biparietal diameter (outer-to-inner), and write output/bpd.json with frame number and BPD in mm using the calibrated pixel spacing.
- Platforms
win· Windowsmac· macOS- Applications
- RadiAnt
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
av· Audio/video over timegeom· Precise geometryvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Frame within +-3 of gold standard plane frames; BPD within 1 mm of gold; measurement calipers in the saved annotation match the reported value. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so frames are reachable only in RadiAnt; reads of the clip by other processes are caught by file-access auditing.
Proposed partial credit
40% correct frame; 50% BPD; 10% annotation consistent
R09.01.01.01-T17 Characterize an adrenal nodule by contrast washout in Horos
Original instruction
In Horos open adrenal-protocol CT ADR-03 (unenhanced, portal venous, 15-min delayed; synthetic). Place matching ROIs over the left adrenal nodule on each phase, compute absolute and relative washout, and write output/adrenal.json with HU values, washouts and whether the result favours adenoma.
- Platforms
mac· macOS- Applications
- Horos
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometrynav· Pan/zoom/scrollvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
ROIs inside the gold nodule mask on all 3 phases covering >=50% of its area; HU within 5 of gold; washout percentages within 5 points; adenoma call exact. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so HU values are reachable only in Horos; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
30% ROI placement; 30% HU values; 25% washout calculations; 15% adenoma call
R09.01.01.01-T18 Localize a breast mass on CC and MLO mammograms in Weasis
Original instruction
In Weasis open screening mammogram MAM-17 (synthetic). Find the mass visible on both right CC and right MLO views, draw an ellipse around it on each view, and write output/mass.json with clock position, distance from nipple in cm, and longest dimension in mm.
- Platforms
win· Windowsmac· macOS- Applications
- Weasis
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometryvis· Visual judgmentnav· Pan/zoom/scroll- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 5 / 5
Proposed verifier
Ellipse IoU>=0.5 with gold on both views; clock position within 1 hour; nipple distance within 1 cm; size within 3 mm. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so images are reachable only in Weasis; reads of the study folder by other processes are caught by file-access auditing.
Proposed partial credit
40% ellipses on both views; 20% clock position; 20% distance; 20% size
R09.01.01.01-T19 Control pair of T03: measure liver and spleen HU with a script
Original instruction
Control pair of T03. Using Python with pydicom and numpy on the same study ABD-102, place 1 cm^2 circular ROIs in the right liver lobe and spleen on one slice level (choose coordinates from the data) and write output/hu.json with both means and the difference.
- Platforms
lin· Linuxmac· macOSwin· Windows- Applications
- Python · pydicom
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
code· Code & commands- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Same thresholds as T03: ROIs (reported centre and radius) inside gold organ masks avoiding vessels; means within 8 HU of gold organ means; enables a GUI-vs-script comparison.
Proposed partial credit
30% valid ROI positions; 40% HU values; 30% difference
R09.01.01.01-T20 Assign Lung-RADS categories to three nodules on a screening CT
Original instruction
In Weasis open low-dose screening CT LDCT-61 and its prior LDCT-61-PRIOR (synthetic). For the 3 nodules listed by location in input/worklist.txt, measure mean diameter, classify as solid/part-solid/ground-glass, assess growth vs prior, and write output/lungrads.json with per-nodule measurements, type and Lung-RADS v2022 category plus overall category.
- Platforms
win· Windowsmac· macOS- Applications
- Weasis
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometrynav· Pan/zoom/scrollvis· Visual judgment- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Per nodule: mean diameter within 1.5 mm, type exact, growth call exact, category exact; overall category exact. Measurements must exist as saved Weasis annotations on the correct slices. The VM has no shell, Python or DICOM toolkits (pydicom, dcmtk), so both studies are reachable only in Weasis; study-folder reads by other processes are caught by file-access auditing.
Proposed partial credit
35% measurements; 20% nodule types; 15% growth; 30% per-nodule and overall categories (2:1)
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.