What it does

Chest CT reads every chest protocol with one model: non-contrast and contrast-enhanced CT, CT pulmonary angiography, low-dose lung-cancer screening and high-resolution CT for interstitial lung disease. The run works out the protocol from the kernel, slice thickness, contrast timing and pulmonary-artery enhancement, then runs the sections that protocol supports.

The result covers the whole chest. Nodules come with size, volume, type and management per the 2017 Fleischner guidelines, or a Lung-RADS v2022 category on a screening scan. On a CTPA, emboli are reported with the most proximal clot level, clot burden and right-ventricular strain. Parenchyma (emphysema, fibrosis, infection), airways, pleura, ribs and coronary calcium complete the read, with lungs, lobes and pulmonary vessels segmented underneath.

Intended use

Decision support for chest CT readers: the on-call resident with a night-time CTPA, the screening-programme radiologist assigning Lung-RADS, the pulmonologist following fibrosis or emphysema. Every finding carries its mask and measurements.

Who it is for

Thoracic and emergency radiologists, pulmonologists, emergency physicians, thoracic surgeons and lung-screening programmes. Developers use the run to add nodule management, PE triage or quantitative lung metrics to a reading platform or registry.

Accepted input

• CT series: CT Image Storage, Enhanced CT or Legacy Converted Enhanced CT, uploaded as a zipped study, via DICOMweb or from a pre-signed URL. NIfTI with explicit spacing is accepted for the segmentation section.
• Several reconstructions of one acquisition: the thinnest lung-kernel series feeds the parenchyma sections and a bone-kernel series the rib sections.
• Prior chest CT in prior_files: nodule growth and the Lung-RADS category of a follow-up screen use it.

Requirements

Requirement	Why
Slices of 2.5 mm or thinner, lung kernel preferred	Nodule detection, volume and type
Pulmonary-artery enhancement on CTPA	A contrast-adequacy check runs before any embolus is called
Prior low-dose scan for a follow-up screen	Lung-RADS growth criteria compare against it; a baseline screen needs none
Inspiratory non-contrast series	LAA-950 is defined on inspiratory, unenhanced lung density
Prone and expiratory HRCT series, when acquired	Prone images separate dependent atelectasis from fibrosis; expiratory images show air trapping
Bone kernel	Rib fracture detection and numbering

Respiratory motion, mixing artefact and screening-dose noise are reported as quality flags.

Optional context

The screening or diagnostic setting and the clinical question (suspected PE, staging, dyspnoea) go in clinical_context. With the detected protocol, they decide between Fleischner and Lung-RADS management.

The result

One sectioned JSON result. Each nodule and clot has its own mask, measurements and lobe location.

Section	Content
Emergency triage	Flags for PE, pneumothorax, aortic dissection or large aneurysm, large pleural or pericardial effusion
Pulmonary embolism	Clot detection, most proximal level (saddle to subsegmental), clot masks and burden, RV-strain flags, contrast adequacy
RV/LV ratio	Maximal RV and LV diameters, the ratio and the main pulmonary-artery diameter
Lung nodules	Detection, 3D segmentation, diameter and volume, solid, part-solid or ground-glass, malignancy score, follow-up
Lung-RADS	Growth against the prior, category, management and the S modifier for significant incidental findings
Lung-cancer risk	Predicted risk of lung cancer over 1 to 6 years from a screening scan
Emphysema	Lung and lobe segmentation, LAA-950 percentage per lobe and distribution
Interstitial lung disease	Reticulation, honeycombing, traction bronchiectasis and ground glass, extent and UIP-pattern category
Infection extent	Consolidation and ground-glass segmentation, percentage involvement per lobe and severity
Airways	Airway tree to sub-segmental level, wall thickness, broncho-arterial ratio, bronchiectasis and mucus plugging
Pleural and pericardial effusion	Segmentation, volume in millilitres and density, with a haemothorax suspicion flag
Pneumothorax	Side, volume and tension signs such as mediastinal shift
Rib fractures	Detected fractures with rib number, type (displaced, non-displaced, buckle, segmental) and flail-segment flag
Coronary calcium	Agatston-equivalent score and ordinal category on non-gated CT, including low-dose scans
Lung, lobe and vessel segmentation	Lungs, five lobes, pulmonary arteries and veins and airways, with volumes
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: Chest CAD SR (TID 4100) with tracking identifiers for nodules; SR TID 1500 Measurement Report; SEG for nodules, clots and lung anatomy; locations coded with CID 6170 Lobe of Lung and CID 6126 Location in Lung.
• Nodules and screening: Fleischner Society 2017 guidelines for incidental nodules; Lung-RADS v2022.
• Embolism: ACR PE-RADS; Qanadli obstruction index; ESC 2019 right-ventricular dysfunction criterion (RV/LV of 1.0 or more).
• Parenchyma: Fleischner glossary 2024 terms; Fleischner 2018 IPF white paper and the ATS/ERS/JRS/ALAT 2022 guideline for UIP patterns.
• Calcium: SCCT/STR 2016 guideline for coronary calcium on non-gated chest CT; CAC-DRS.

Limitations and failure modes

• Sections in version 1.0.0: pulmonary embolism (with the RV/LV 1.0 flag). Nodules, effusions, pneumothorax, rib fractures, parenchyma and the RV/LV ratio value are listed as planned.
• Clot level is given as central (main or central pulmonary arteries) or peripheral; lobar, segmental and subsegmental are not separated in this version.
• Small peripheral clots (under about 100 mm3) are outlined less often than larger ones; the exam call also uses the exam-level ensemble, so a study can be called positive with fewer clots outlined than present.
• Degraded studies (respiratory motion, contrast mixing, streak artefact) produce most false calls; the opacification check catches poor contrast, not motion.
• Chronic emboli are reported as emboli; acute and chronic are not separated.
• RV/LV is a flag (ratio 1.0 or more), not a measured ratio; ventricle diameters join with the rv-lv-ratio section.
• Non-CTPA phases: venous-phase and non-contrast studies are stated not assessable for embolism rather than read.