Riverain Technologies

The software is the product. Riverain describes ClearRead as AI powered tools for chest CT and chest radiographs, and the ClearRead CT 510(k) summary (K161201) describes a post-processing application that generates a secondary, vessel suppressed lung CT series with detection marks and region descriptors.

Vessel suppression is the technical thesis: rather than only marking findings, the software produces an altered image in which the vessels that can hide nodules are suppressed, read alongside the original series. The company sells no scanner or services layer, so without the models there is nothing left to sell.

ClearRead CT is cleared as computer assisted reading tools that aid the radiologist in detecting pulmonary nodules, and its 510(k) summary (K161201, September 2016) states that it provides adjunctive information and is not intended to be used without the original CT series. The radiologist reads the original images and makes the call; the vessel suppressed series and the detection marks support that read.

The parts of the oversight structure the summary does not state are the reading mode, meaning whether the output is meant to be viewed during the radiologist's first read or only after it, which changes how far the marks can steer where the radiologist looks, and the operating point at which a mark is shown.

The approach is named. The ClearRead CT 510(k) summary (K161201) describes a post-processing application that suppresses vessels in the lung CT to produce a secondary series and places detection marks with region descriptors on it, and Riverain describes the ClearRead products as AI powered.

ClearRead CT clearance documentation reports performance estimates by nodule type, a residual analysis showing little effect across slice thickness, normalization of scans across manufacturers and acquisition protocols without per device tuning, and a micro nodule filtering option whose site configurable threshold is capped at 5 mm to keep behavior within labeled performance. The model architecture, the scale and composition of the training data, and the versioning and update practice are not published.

The chain is short at the model layer. ClearRead CT is Riverain's own device, cleared under the company's name (K161201), so the models producing the output are the company's own cleared models rather than a third party's. Who else handles studies is not named: no hosting provider, cloud service or subprocessor is named for the products, and whether studies leave the hospital at all depends on how ClearRead is deployed, which the home page does not state. The privacy policy covers the website and email only; it names Google Analytics and unnamed service providers for those, and says that information may be processed in the United States and elsewhere.

The core evidence is the multi reader multi case study behind the 2016 clearance. The 510(k) summary (K161201) states that the study measured the area under the localization ROC curve and reading time with ClearRead CT against the unaided read, and that the device significantly increased the area under the curve and decreased reading times; it gives no figures, reader numbers or case counts. Riverain reports the study as showing a 29 percent reduction in missed actionable nodules.

An independent study led by University of Washington School of Medicine radiology researchers examined the vessel suppression product. The home page reports more than 400 customer locations, which is scale of use rather than evidence of benefit. Outcome level data, such as effects on stage at diagnosis, is not published.

There is no published account of how ClearRead handles the studies it processes. The K161201 summary describes a postprocessing application that generates a secondary series from the lung CT; where that processing runs, what is retained afterward, whether customer studies are used to train or improve the models, and whether any de identification is applied are not stated.

The privacy policy, last modified in March 2026, applies to information collected on the website or in emails and does not address studies or outputs, and the site's footer links no other privacy, security or data page. Safety engineering, meaning guardrails and how a safety event is handled, is not published either; the labeling that requires the original series is an oversight control.

Riverain states no business associate status, and no privacy document that reaches the product is published. The privacy policy, last modified in March 2026, applies to information collected on the website or in emails and does not mention HIPAA or protected health information, and the site's footer links no other privacy, security or compliance page.

A US hospital using ClearRead is the covered entity; whether Riverain is its business associate turns on whether studies reach Riverain or a host acting for it, which depends on a deployment model the home page does not state.

No information security certification of Riverain's own is published, and the site's footer links no security or trust page. The privacy policy says the company uses reasonable technical, administrative and physical safeguards and cannot promise complete security, which asserts controls with nothing independent behind them.

ISO 13485 and the Medical Device Single Audit Program, where cited for the company, are quality management audits of how the device is designed and made; neither examines encryption, access control, monitoring or incident response. What a breach at Riverain could expose depends on whether studies ever leave the hospital, which is not stated.

The regulatory position is identifiable in the FDA database. ClearRead CT was cleared on 9 September 2016 (K161201, product code OEB, 21 CFR 892.2050) as computer assisted reading tools that aid the radiologist in detecting pulmonary nodules during review of chest CT in an asymptomatic population; both lungs must be in the field of view, and the output is adjunctive and not for use without the original series.

Later clearances under Riverain Technologies include ClearRead Xray Pneumothorax (K213566, 10 March 2022, product code QFM, 21 CFR 892.2080) and ClearRead CT CAC for coronary artery calcium (K242188, 3 December 2024, product code JAK, 21 CFR 892.1750). Earlier ones include ClearRead +Confirm (K123526, December 2012) and DeltaView 2.1 (K111776, December 2011, filed as Riverain Medical Group), and K161201 names ClearRead Bone Suppression (K092363) as a predicate. The home page also lists nodule detection on chest radiographs under ClearRead Xray; the clearances above do not include it, and its US status is not stated on the home page.

Riverain publishes no results for its detection function compared across patient or nodule groups, and no governance framework or post market monitoring commitment. The original ClearRead CT 510(k) summary (K161201, September 2016) reports its reader study with no subgroup analysis.

The ClearRead CT Compare summary (K221612, December 2022) reports nodule matching rates by nodule type (0.961 solid, 0.957 part solid, 0.946 ground glass) and by location (0.940 isolated, 0.969 juxta vascular, 0.955 juxta pleural) on 900 nodules with confidence intervals, and says performance across slice thicknesses of 3.5 to 5.0 mm changed little.

Those figures describe the matching step, which the filing describes as volume registration and does not describe as learned, so they do not show how the AI the company sells behaves across groups. The company's Sectra Amplifier Marketplace listing says the product is effective across all nodule types (solid, part solid and ground glass), a robustness statement with no detection results or counts behind it. Nothing is published on performance by patient sex, age or body habitus.

Riverain publishes no warranty, indemnity or remediation commitment, and neither of its ClearRead CT 510(k) summaries gives an error rate for a learned function. The first (K161201, September 2016) reports a multi reader multi case study in which the device increased the area under the localization ROC curve and decreased reading times, with no standalone sensitivity, false positive rate or case counts.

The second (K221612, 5 December 2022, Riverain Technologies, Miamisburg, Ohio) adds ClearRead CT Compare, which aligns current and prior scans with a volume registration step to match nodules and report volume change and doubling time.

Its standalone tests report nodule matching rates on 900 nodules (0.946 to 0.961 by nodule type, 0.940 to 0.969 by location) and a 25 patient clinical cohort in which all 42 actionable nodules ground truthed by a radiologist were detected and correctly matched or identified as new, with an average registration error of 4.46 mm; false positives are not counted.

Neither summary names a machine learning component or says which outputs are learned, while the company describes its products as AI powered, so these figures measure the matching and registration the filing describes rather than an error rate for the AI the product is sold on. The labeling makes ClearRead CT adjunctive and requires the original series, so the radiologist's read of the original images is the check on a missed nodule or a false mark.

ClearRead works in the imaging workflow rather than the record system. Its listing on the Sectra Amplifier Marketplace, the application store for Sectra's PACS, says output is provided in DICOM and readable in the PACS, with regions of interest and measurements exported as an overlay object or a DICOM secondary capture object, the vessel suppressed series available alongside the original, an optional worklist and pre populated reports, and tells buyers to check which Sectra PACS version is compatible.

The 510(k) summaries describe delivery through the same DICOM network interface that supplies the CT series. The home page displays the names of health systems including the VA, Cleveland Clinic, Duke, Cedars-Sinai, Michigan Medicine and University Hospital Zurich, and reports more than 400 customer locations. That is a named archive receiving one directional output. No EHR integration is claimed, and no HL7 or FHIR path for results is published.

Where ClearRead runs is not committed on the home page or in the privacy policy. Riverain describes ClearRead CT as installing into the existing clinical environment without new hardware and without per device tuning, which suggests processing inside the hospital; that is an inference, since no on premises or hosted option is stated. The privacy policy covers only the website and says website information may be processed in the United States and elsewhere, which says nothing about studies. Tenancy, hosting and residency terms for the product are not published.

No price, pricing basis or unit of charge is published. One route to buy is public: ClearRead CT is listed as an application in the Sectra Amplifier Marketplace, the application store for Sectra's PACS, with its integration and regulatory details and no price or licensing terms. The company's home page gives no licensing basis, whether per site, per study or across a health network.

Coverage is named and deliberately cardiothoracic: chest CT for pulmonary nodules and coronary artery calcium, and chest radiographs for bone suppression, nodule detection and pneumothorax. Validation sits behind part of it. The nodule detection clearance (K161201) covers chest CT in an asymptomatic population with both lungs in the field of view, supported by a multi reader study, so use on diagnostic chest CT in symptomatic patients falls outside the indication as summarized.

Calcium scoring (K242188) and pneumothorax (K213566) are cleared; radiograph nodule detection has no US clearance stated. ClearRead CT LTA is listed as research use only. Coverage does not extend beyond the chest.