Gleamer

The models are the product. Gleamer sells a suite of detection, measurement and reporting models under the Gleamer Copilot name: BoneView for fractures on radiographs, ChestView for chest X-ray findings, BoneMetrics for musculoskeletal measurements, BoneAge for bone maturity, LungCT for pulmonary nodules, BoneCT for secondary bone lesions, BreastView for breast lesions, the Pixyl neuro products for multiple sclerosis and brain volumetry, LumbarMR for spine pathologies, and two reporting products, Voice, described as an AI reporting agent, and AutoReport.

There is no hardware, archive or services layer underneath; remove the models and nothing is left to sell. Gleamer has been part of DeepHealth, the digital health subsidiary of RadNet, since March 2026, and its site and products still carry the Gleamer name.

BoneView is a concurrent reading aid: the clinician reads every radiograph and makes the call. Its FDA 510(k) summary for BoneView 1.1-US (K222176, March 2023) describes four outputs per study: a solid box for a high confidence fracture (FRACT), a dotted box for a suspicious or subtle finding (DOUBT FRACT), no fracture, and not available when a study falls outside what the software can analyze. The summary publishes the two operating points behind the boxes with their sensitivity and specificity, restricts use in children and adolescents aged 2 to 21 to the limbs, and states prescription use only. That is a described oversight structure with its thresholds published.

What is missing is what happens after an error: nothing describes how a missed fracture or a false box is reported, reviewed or fed back to the company. Gleamer's home page also lists two newer reporting products, Voice, described as an AI reporting agent, and AutoReport, and how a clinician reviews a drafted report before sign off is not described there.

The FDA 510(k) summary for BoneView 1.1-US (K222176, March 2023) identifies the approach and characterizes the training data: supervised deep learning, trained on 44,649 radiographs (151,096 images) covering all anatomical areas from multiple equipment manufacturers, 52.4 percent male, aged 0 to 109. It names the version and states that the test data were independent of the data used for training, tuning and setting the operating points. What is not published is update practice.

The summary carries no predetermined change control plan, and nothing states how and when models are retrained or how a customer learns that a new version is running. The ChestView and BoneMetrics summaries (K241620, K252486 and K241593) are the equivalent documents for those products.

The model layer is Gleamer's own. BoneView, ChestView and BoneMetrics are FDA cleared devices in Gleamer's name (K222176, K241620, K252486 and K241593), and BoneView's summary describes a supervised deep learning model trained by the company, so the models cannot be swapped for another supplier's without a new submission. That keeps the chain structurally short at the model layer. The rest of the chain is partly stated.

Gleamer's privacy policy, which covers its products as well as its website, says it engages subprocessors for services such as data hosting and database administration, without naming them, and states that no transfer is made outside the European Union. The hosting provider, the subprocessor list and the path taken by United States customers' studies are not published, and since March 2026 Gleamer has been part of DeepHealth, whose infrastructure may enter the path.

A fully crossed multiple reader, multiple case study is reported in BoneView's FDA 510(k) summary (K222176, March 2023): 24 clinical readers, 480 cases, randomized aided and unaided reads with a washout period, and a reference standard set by a panel of three US board certified radiologists. Aided readers' sensitivity rose from 0.648 to 0.752 and specificity from 0.906 to 0.956, and aided performance was higher in every anatomical region. Standalone tests on independent data cover 8,918 adult radiographs (3,886 with a fracture) and 2,000 pediatric radiographs (1,000 with a fracture). These are retrospective test set results: they show what the aid does on curated cases, not in live reading.

The company reports more than 140 clinical studies and more than 60 publications, and its clinical studies library lists a systematic review of BoneView and emergency department studies. Independent evaluations of commercial fracture detection algorithms on real world emergency department data have also been published, though it is not clear whether they cover the BoneView version sold. No prospective, multisite evaluation of BoneView in live use is published. The company's 70 percent reduction in reading time, cited for BoneMetrics with automated reporting, has no method or denominator beside it.

The privacy policy covers Gleamer's products as well as its website, and it is specific about deidentification and training use. When Gleamer receives patient data from a customer, including imaging examinations and reports, it acts as the customer's processor and works on pseudonymized data, and where the customer authorizes it, Gleamer may anonymize pseudonymized patient data and use the anonymized result for its own purposes, including training and developing its products.

On retention it gives no window for imaging data, only that personal data is kept no longer than necessary for its purposes. That is a specific, checkable training position without a retention schedule behind it. Safety engineering, meaning guardrails and how a safety event is handled, is not published. The confidence outputs and prescription only labeling are oversight controls, which is a different thing.

Gleamer publishes a substantive privacy document that reaches the service, and no business associate status. Its privacy policy, updated in 2026, covers its products and states that when it receives patient data, including imaging examinations and reports, the customer is the controller and Gleamer is the processor, with a data protection officer named. That is a European data protection position.

Gleamer also sells into the United States under FDA clearance, and its one United States statement is a line in the BoneView 510(k) summary (K222176) marking the device HIPAA compliant. No business associate status, agreement, tier or subcontractor terms are published. A service privacy document with no business associate status stated, beside a bare compliance claim, grades here. Since March 2026 Gleamer has been part of DeepHealth, a United States company, so the signing entity is an open question.

Gleamer publishes no information security certification and no outside check of its security. The site footer carries privacy, terms, cookie and voice product policies and no security or trust page. The privacy policy asserts technical and organizational measures and names ISO 13485:2016, which certifies the quality management system for designing and manufacturing medical devices, not how the company operates information security, and FDA clearance reviews the device rather than the vendor's operations.

Gleamer has been part of DeepHealth, the digital health subsidiary of RadNet, since March 2026, and DeepHealth's own site carries no security, trust or compliance page either: its footer links accessibility, privacy, terms, a carbon reduction plan and a modern slavery statement, and its page sitemap lists privacy notices, terms and a meaningful use certification page, with nothing on a security attestation or on whether one extends to Gleamer's products. Controls asserted with nothing independent behind them grade here.

Five FDA 510(k) clearances are listed under the applicant Gleamer, all Class II radiology software. BoneView: K212365 (1 March 2022, 21 CFR 892.2090, product code QBS) for adults, and K222176, BoneView 1.1-US (2 March 2023), which added children and adolescents aged 2 to 21 for the limbs (ankle, foot, knee, tibia and fibula, wrist, hand, elbow, forearm, humerus, shoulder and clavicle) while keeping all regions, including pelvis, ribs and spine, for adults over 21.

BoneView is a concurrent reading aid for 2D radiographs, for prescription use only. The pediatric extension was cleared in March 2023. BoneMetrics (US): K241593 (5 February 2025, 21 CFR 892.2050, QIH). ChestView US: K241620 (27 February 2025, 21 CFR 892.2070, MYN) and ChestView US 1.1, K252486 (9 September 2026). No United States clearance under Gleamer is listed for BoneAge, LungCT, BoneCT, BreastView, the Pixyl neuro products, LumbarMR, Voice or AutoReport, and their US regulatory status is not stated.

BoneView's FDA 510(k) summary (K222176, March 2023) publishes results compared across groups the product serves, with numbers. By age: at the high sensitivity operating point, sensitivity 0.928 and specificity 0.811 on 8,918 adult radiographs against 0.909 and 0.821 on 2,000 radiographs of children and adolescents; at the high specificity point, 0.841 and 0.932 in adults against 0.792 and 0.965 in children.

By anatomical region, sensitivity for children ranged from 0.852 to 1.0 and specificity from 0.667 to 0.928, so a buyer can see where the model underperforms, and region results are tabulated for adults as well. Test data came from nine equipment manufacturers and were independent of the training data. What is not published is performance by sex or by equipment manufacturer, which the summary describes but does not break out. The ChestView and BoneMetrics summaries (K241620, K252486 and K241593) are where the equivalent results for those products would sit.

BoneView's error rate is on the record in the manufacturer's own filing. The FDA 510(k) summary for BoneView 1.1-US (K222176, March 2023), which Gleamer prepared, reports standalone performance of the supervised deep learning model on data independent of training, at the two operating points the product is sold with.

On 8,918 adult radiographs (3,886 with a fracture), sensitivity was 0.928 and specificity 0.811 at the high sensitivity point, and 0.841 and 0.932 at the high specificity point. On 2,000 radiographs of children and adolescents (1,000 with a fracture), the figures were 0.909 and 0.821, and 0.792 and 0.965. The summary does not state who set the reference standard for these standalone tests. These are test set results for the cleared indications, not performance in use, and they cover BoneView only. No warranty, indemnity or remediation commitment attaches to the output.

BoneView takes DICOM images from X-ray systems or the archive and returns annotated result images and a summary table into the reading workflow, on premises or in the cloud, according to its FDA 510(k) summary (K222176). Named sites use the products in production: the home page quotes clinicians at Caritas Hospital Berlin and St Luke's. That is standards support with named deployments behind it.

No integration with a named record system, RIS or archive vendor is published, and no HL7 or FHIR path for results is described. Gleamer has been part of DeepHealth, the digital health subsidiary of RadNet, since March 2026, and nothing published says whether the products now run inside DeepHealth's platform or remain available on other vendors' archives.

Deployment options and residency are stated; isolation and the processing path are not. BoneView runs on premises or in the cloud, according to its FDA 510(k) summary (K222176). Gleamer's privacy policy, which covers its products, states that no transfer is made outside the European Union and that the company uses subprocessors for data hosting, without naming them.

What is open is the hosting provider and region for the cloud option, tenant isolation, and how the European residency statement applies to United States customers' studies. Since March 2026 Gleamer has been part of DeepHealth, a United States company, which may change the answer over time.

No price, unit of charge or tier is published. The packaging and the owner are public. The products are sold as modules of the Gleamer Copilot suite, and since March 2026 Gleamer has been part of DeepHealth, the digital health subsidiary of RadNet, so contracting may run through DeepHealth. Neither fact sizes the cost. The company's scale figures, more than 60 million exams a year across more than 3,500 clinical sites in more than 50 countries, describe use, not price.

Coverage is named, with validation behind the lead product. BoneView is cleared for adults across all anatomical regions, including pelvis, ribs and spine, and for children and adolescents aged 2 to 21 on the limbs only, for use by clinicians in primary care, emergency medicine, specialty care and radiology; its FDA 510(k) summary (K222176) validates both populations on independent test sets and adults in a reader study.

The suite now claims far wider coverage, described by the company as covering all imaging modalities, with products for chest X-ray, chest CT, bone CT, breast imaging, brain MRI and lumbar spine MRI. Validation for those is not summarized on the home page, and ChestView and BoneMetrics have their own FDA summaries. BoneView's summary validates on test sets rather than by care setting, so performance in each of the named settings is not shown.