Radiology AI medical devices

A neutral, evidence-based catalogue of radiology AI medical devices. No manufacturer pricing, no buy links — just regulatory facts and clinical capability summaries.

Recommended

Radiology

Aidoc

Aidoc Medical

CEFDATGA

Verified clinician reviews launching soon

Multi-condition CT triage platform (aiOS™) covering intracranial haemorrhage, pulmonary embolism, aortic dissection, large and medium vessel occlusion, cervical-spine fracture, rib fracture, pneumothorax, and further abdominal findings — surfaced as prioritised worklist flags to radiologists and emergency clinicians in real time.

Exceptional

Radiology

Transpara

ScreenPoint Medical B.V.

CEFDAHCTGA

Verified clinician reviews launching soon

Deep-learning reading support for breast-cancer screening. Transpara analyses full-field digital mammography and digital breast tomosynthesis exams, marking calcification groups and soft-tissue regions with per-region malignancy scores and assigning an exam-level score used both as concurrent decision support and — in European screening programmes — to triage exams between single and double reading. It is the mammography AI with the strongest trial evidence on the market: the 105,000-woman MASAI randomised trial, run independently of the manufacturer, reported more cancers found, fewer interval cancers missed, and a 44% cut in screen-reading workload.

Recommended

Radiology

ProFound AI for Digital Breast Tomosynthesis

iCAD (DeepHealth / RadNet)

CEFDAHC

Verified clinician reviews launching soon

Concurrent-read computer-aided detection for digital breast tomosynthesis (DBT). ProFound analyses every reconstructed tomosynthesis plane, marks suspicious soft-tissue densities and calcifications with a per-finding Certainty of Finding score, and assigns a per-case score reflecting the algorithm's overall suspicion for the study. It is an assistive overlay read alongside the radiologist, not an autonomous or triage device.

Recommended

Radiology

Viz.ai

CEFDA

Verified clinician reviews launching soon

Care-coordination platform (Viz.ai One) that analyses imaging and other signals for time-critical findings — most prominently large-vessel occlusion (LVO) on CT angiography — and alerts the appropriate specialist on a mobile application in parallel to standard image interpretation. The suite has since expanded across stroke (LVO, perfusion, intracranial and subdural haemorrhage, aneurysm), vascular (abdominal aortic aneurysm, pulmonary embolism), and cardiac (AI-enabled ECG screening for hypertrophic cardiomyopathy) pathways.

Recommended

Radiology

Annalise CXR

Annalise.ai

CEFDATGA

Verified clinician reviews launching soon

Comprehensive chest X-ray AI that classifies up to 124 clinical findings simultaneously, with heatmap localisation and a radiologist-style structured report in its CE-marked decision-support form, and a five-finding triage-and-notification variant (Annalise Triage CXR) cleared in the United States for pneumothorax, tension pneumothorax, pleural effusion, pneumoperitoneum, and vertebral compression fracture.

Exceptional

Radiology

qXR

Qure.ai

CEFDAHCTGA

Verified clinician reviews launching soon

A deep-learning chest X-ray platform whose scope depends sharply on jurisdiction. The international, CE-marked product reads frontal chest radiographs for roughly 29-30 abnormality categories with heatmap localisation and a structured report. In the United States the brand exists only as a set of narrow, single-purpose clearances — breathing-tube position, pneumothorax/effusion triage, cardiothoracic-ratio measurement, lung-nodule computer-aided detection, and (from 2026) a multi-finding CADe device, qXR-Detect — each cleared for one task rather than the broad reporting tool sold elsewhere.

Exceptional

Radiology

Lunit INSIGHT MMG

Lunit Inc.

CEFDAHCUKCA

Verified clinician reviews launching soon

Deep-learning reading support for breast-cancer screening. Lunit INSIGHT MMG analyses full-field digital mammography exams, marking regions suspicious for breast cancer with abnormality scores to support the interpreting radiologist. It carries an unusual distinction in this catalogue: it is the system a Swedish screening hospital used to become the first in the world to replace one of two human readers with AI in routine double reading, on the strength of the prospective ScreenTrustCAD trial — 55,581 women, AI plus one radiologist non-inferior to two radiologists — published in The Lancet Digital Health in 2023.

Exceptional

Radiology

Lunit INSIGHT CXR

Lunit Inc.

CEFDAHCUKCA

Verified clinician reviews launching soon

Deep-learning software that reads frontal chest radiographs and flags suspected thoracic abnormalities — up to ten findings in the CXR3 line (nodule, consolidation, pneumothorax, pleural effusion, cardiomegaly, fibrosis and others), extended to twelve including acute bone fracture in CXR4 — returning a per-finding abnormality score and a heat-map or outline overlay. It runs as a concurrent or second reader; the radiologist reviews every case. A separate, narrower product — Lunit INSIGHT CXR Triage — is what carries the US FDA clearance, and it notifies only on two critical findings rather than detecting the full panel.

Recommended

Radiology

SubtleMR

Subtle Medical

CEFDA

Verified clinician reviews launching soon

Vendor-agnostic, DICOM-based deep-learning software that denoises and sharpens magnetic-resonance images after acquisition. By restoring signal-to-noise and apparent resolution on faster, lower-resolution pulse sequences, SubtleMR lets sites shorten scan time — the manufacturer describes accelerations of roughly a third to a half — while aiming to preserve the diagnostic quality of the standard-of-care study. It is an image-processing overlay, not a detection or diagnostic tool: it enhances the pictures a radiologist reads, and the radiologist remains the reader.

Recommended

Radiology

AI-Rad Companion Chest CT

Siemens Healthineers

CEFDAPMDA

Verified clinician reviews launching soon

Multi-organ post-processing AI for chest CT. One automated pass segments the lungs and lung lobes, detects and volumetrises pulmonary nodules, measures aortic diameters at guideline landmarks, quantifies heart volume and coronary calcium, and measures vertebral body heights and density — returning the results as a structured report and overlay series alongside the standard radiologist read. Every US clearance in the chain is quantification software, not computer-aided detection.

Exceptional

Radiology

ClearRead Xray

Riverain Technologies

CEFDAHCTGA

Verified clinician reviews launching soon

A family of chest-radiograph processing modules built around Riverain's single-exposure bone-suppression technology. Bone Suppress produces a soft-tissue image from a standard chest X-ray without dual-energy hardware; Detect (a PMA-lineage CAD) marks suspected pulmonary nodules; Confirm and Enhance sharpen tube-and-line and general visibility; Compare registers current and prior films to highlight change; and the 2022-cleared Pneumothorax module provides notification-only triage for suspected pneumothorax of 5 mm or larger. All modules are adjunctive — the radiologist reads the original image and makes every call.

Exceptional

Radiology

ClearRead CT

Riverain Technologies

CEFDAHCTGA

Verified clinician reviews launching soon

A suite of machine-learning image-processing modules for chest CT. The core capability is vessel suppression — the software subtracts pulmonary blood vessels from the CT volume so that small nodules, which are easily hidden against the vascular tree, stand out — paired with a computer-aided detection (Detect) module that marks candidate nodules for the reading radiologist. Later modules add longitudinal comparison of prior and current studies (Compare) and, in a separately cleared product, automated coronary-artery-calcium Agatston scoring on ungated chest CT (ClearRead CT CAC). Every module is adjunctive and concurrent-read: the software processes the original CT and presents suppressed images and marks alongside it, and the radiologist makes the diagnosis.

Exceptional

Radiology

Rapid

iSchemaView, Inc. (d/b/a RapidAI)

FDAHCPMDATGA

Verified clinician reviews launching soon

Neuroimaging software that processes CT perfusion, CT angiography, MRI and non-contrast CT for time-critical stroke findings and, in its computer-aided triage modules, alerts a stroke specialist when a suspected large-vessel occlusion, intracranial haemorrhage, subdural haemorrhage or aneurysm is detected. The platform grew out of the automated perfusion-processing engine that quantifies ischaemic core and salvageable penumbra — the imaging measure used to select patients for late-window thrombectomy in the DAWN and DEFUSE 3 trials — and has since expanded into a family of FDA-cleared detection, triage and quantification modules across the acute-stroke pathway.

Recommended

Radiology

LiverMultiScan

Perspectum Ltd (formerly Perspectum Diagnostics Ltd), Oxford, UK

CEFDA

Verified clinician reviews launching soon

A post-processing software device that turns a short, non-contrast MRI of the liver into standardised quantitative maps of tissue health. From T1 mapping, proton-density fat-fraction and T2* sequences it derives three biomarkers: iron-corrected T1 (cT1), a composite marker of fibro-inflammation; liver fat (PDFF); and iron (T2*/R2*). Its analysis pipeline places and quality-controls regions of interest and computes the metrics against reference ranges, so the interesting boundary to hold in view is what kind of tool this is: it is a quantification aid, not an autonomous or diagnostic reader. The FDA and CE clearances cover measurement of liver-tissue properties to support a clinician's interpretation; the software does not stage disease, make a diagnosis, or recommend a course of action on its own. It appears on the FDA's list of AI/ML-enabled devices, but its core is physics-based mapping with software-assisted analysis rather than a learned diagnostic classifier.

Exceptional

Radiology

Optellum Virtual Nodule Clinic

Optellum Ltd.

CEFDATGAUKCA

Verified clinician reviews launching soon

Computer-assisted diagnostic software for indeterminate pulmonary nodules on chest CT. Its Lung Cancer Prediction score, produced by a convolutional neural network trained on screen-detected and incidentally found nodules, gives radiologists and pulmonologists a calibrated malignancy probability to inform whether a nodule is watched, biopsied or resected. It was the first CADx software of its kind cleared under all three of FDA 510(k), EU MDR and UKCA, and the evidence for the score's discrimination is published in radiology and respiratory journals.

Approved

Radiology

MammoScreen

Therapixel SA

FDA

Verified clinician reviews launching soon

Deep-learning reading aid for breast-cancer screening. MammoScreen analyses full-field digital mammography and digital breast tomosynthesis exams, marking soft-tissue lesions and calcification groups and assigning each a suspicion score — rolled up per breast and per exam on the 1–10 MammoScreen Score scale. The current generation compares against prior exams and wraps the detection core in workflow features: a worklist complexity index, case pre-check and automated pre-reporting. The algorithm's lineage is unusual — it grew out of Therapixel's winning entry in the 2017 Digital Mammography DREAM Challenge.

Recommended

Radiology

Koios DS

Koios Medical, Inc.

CEFDAHC

Verified clinician reviews launching soon

Machine-learning decision-support software for diagnostic ultrasound. Koios DS takes a physician-selected region of interest on a breast or thyroid ultrasound and returns a probability-of-malignancy read mapped to the standard reporting lexicons — the four ACR BI-RADS assessment bands for breast lesions, and ACR TI-RADS with American Thyroid Association descriptors for thyroid nodules. It is an adjunct, not an autonomous reader: the interpreting physician selects the lesion, reviews the software's descriptors and suspicion read, and makes the management call. The breast module descends from ClearView Diagnostics' morphology classifier; thyroid analysis was added with the 2021 clearance.

Approved

Radiology

contextflow ADVANCE Chest CT

contextflow GmbH

Verified clinician reviews launching soon

A chest-CT decision-support suite that pairs content-based image retrieval with detection and quantification. Its original module — first marketed as SEARCH Lung CT — lets a radiologist select a region of interest and retrieve visually similar cases from a reference library of thousands of expert-labelled thin-section CT scans, surfacing the differential diagnoses associated with matching imaging patterns. Around that retrieval core the product adds detection of lung nodules and incidental pulmonary embolism, quantification of nineteen lung-tissue patterns relevant to interstitial lung disease and COPD (including emphysema), longitudinal nodule tracking, and — through a partnership with RevealDx — a Malignancy Similarity Index that scores detected nodules. In every module the radiologist remains the reader; the software is concurrent decision support, not an autonomous diagnosis.

Recommended

Radiology

Veye Lung Nodules

Aidence B.V. (DeepHealth, a RadNet company)

CEFDA

Verified clinician reviews launching soon

Deep-learning software that automatically detects, measures, classifies and tracks pulmonary nodules on chest CT, including volume and volume-doubling time across serial scans. Veye runs as a concurrent or second reader — the radiologist reviews every finding — and a reporting add-on prepopulates the structured protocol used by the NHS England Targeted Lung Health Checks, where it is the lung-screening AI of choice. CE-marked Class IIb under the EU MDR; not FDA-cleared, and not distributed in the United States.

Radiology