Qure.ai and Harrison.ai address different parts of the radiology AI procurement problem. Qure.ai's public materials center on specific imaging products, particularly qXR for chest X-ray reporting, with US FDA 510(k) clearances and documented output formats. Harrison.ai's public materials describe a broader radiology platform, a partner ecosystem, and a foundation model that is explicitly not cleared for clinical use. A buyer comparing the two is not choosing between equivalent chest X-ray algorithms. The comparison turns on whether the organization wants a focused, regulated detection and reporting tool or a platform intended to host multiple algorithms, including third-party products, through one integration. Several important details remain unresolved from the available documentation, including exact deployment requirements and comparative performance in the same clinical setting.

Documented Workflow and Intended Setting

Qure.ai's qXR is presented as a chest X-ray interpretation tool that can provide pre-read assistance, pre-filled reports, and AI overlays [1]. The Qure.ai App user manual describes an on-cloud image viewing workflow in which DICOM studies arrive from a PACS or modality, are processed on Qure's cloud server, and AI-processed reports are displayed in the app [7]. The manual states that reports can be read, viewed, and shared with radiologists, consultants, and doctors [7]. This positions qXR within a reporting workflow where the AI output becomes part of the radiologist's review process, rather than a standalone diagnostic replacement.

Harrison.ai's public materials describe a different organizational setting. The Harrison.ai Open Platform, or HOP, is framed as a single integration with PACS or RIS through a component called the Harrison Orchestrator [9]. The vendor states that the platform provides access to Harrison.ai algorithms and partner algorithms under one roof, with per-study pricing and no platform markup [9]. Harrison.ai also markets Comprehensive Care, described as a triage solution for chest X-rays and non-contrast head CT that surfaces findings within the existing workflow [11]. The intended setting is therefore a radiology group or hospital seeking to consolidate multiple AI tools, while Qure.ai's qXR documentation centers on a specific chest X-ray reporting task.

Regulatory Scope and US Indications

Qure.ai's US materials state that qXR-LN is FDA cleared for radiologists, pulmonologists, and emergency physicians reviewing adult chest imaging [2]. The vendor also states that qCT LN Quant is FDA cleared for radiologists and pulmonologists [2]. A February 2026 announcement describes FDA 510(k) Class II clearance for qXR-Detect, a computer-assisted detection device for plain film chest X-rays intended to support emergency physicians, family medicine practitioners, and radiologists [8]. The announcement states that Qure.ai's FDA cleared indications total 26 across nine products for X-ray and CT [8]. These are vendor statements about regulatory status, not independent verification of clinical performance.

Harrison.ai's US materials describe Comprehensive Care as having 13 FDA-cleared findings, including four intracranial hemorrhage types, obstructive hydrocephalus, acute infarct, vasogenic edema, pneumothorax, tension pneumothorax, pleural effusion, mass effect, and pneumoperitoneum [11]. The vendor states that this is the marketing name for Annalise Critical Care AI and that it is intended to assist clinicians with triage and notification of suspected findings [11]. Harrison.ai's global radiology solutions page lists CE mark Class IIb status for non-contrast head CT and CT chest products [12]. The regulatory claims for the two vendors are not directly comparable because they cover different modalities and finding sets.

A Suggested Evaluation Workflow

A radiology group evaluating both vendors could begin by defining the primary workflow. If the need is chest X-ray pre-read assistance with structured outputs, the group could request a qXR API test using the documented input criteria and output formats [4][6]. If the need is a multi-algorithm platform, the group could ask Harrison.ai to demonstrate HOP integration with the existing PACS and a subset of partner algorithms [9]. This suggestion is an evaluation approach, not a claim about which vendor performs better. The group should also request current regulatory documentation for the specific products and regions under consideration, since Harrison.ai's global materials state that product availability and regulatory status vary by country [12].

Technical Documentation and Implementation Boundaries

Qure.ai publishes platform API documentation that specifies output formats for qXR. The documented outputs include GSPS overlays, PDF reports, DICOM encapsulated reports, structured reports, and secondary capture files [4]. The API documentation states that generated outputs are available as a response payload with URLs for file downloads [5]. A separate qXR input specification lists acceptance criteria such as PA or AP view, minimum resolution of 1440 by 1440, patient age over six years, erect or supine position, and valid DICOM format [6]. These boundaries are useful for integration teams because they define what the algorithm expects and what it returns.

Harrison.ai's technical documentation is less focused on a single product's input and output contract. The Harrison.Rad 1.5 technical report describes a foundation model trained on approximately 6.5 million studies, yielding around 18 million conversations [10]. The report states that the model has not been cleared for clinical use and is not a medical device [10]. It also notes that the model is trained exclusively on 2D X-ray modalities and that 3D volumetric imaging support is planned for future iterations [10]. This documentation is a research preview, not an implementation guide for a regulated clinical product. Buyers evaluating Harrison.ai for production use must rely on separate product documentation for Comprehensive Care or HOP.

Customer Segment Differences

Qure.ai's qXR materials emphasize high-volume chest X-ray settings, including teleradiology and immigration screening [1]. The vendor cites a case study with Aarthi Scans in India describing a 40 percent turnaround time reduction by classifying 40 percent of normal X-rays with a stated 98.9 percent negative predictive value [1]. The US site also describes support for radiologists, pulmonologists, and emergency physicians across the lung nodule pathway [2]. These examples suggest a customer base that includes teleradiology groups and hospital imaging departments with substantial chest X-ray volume.

Harrison.ai's materials target radiology groups and hospitals that want to consolidate multiple AI algorithms. The HOP page lists partner vendors including AZmed, Radiobotics, Therapixel, Lucida Medical, Nicolab, Us2.ai, CoLumbo, Lunit, Nanox.AI, and Koios Medical [9]. The vendor states that HOP provides a complete AI suite under one roof and reduces IT integration costs [9]. Harrison.ai also states that its solutions are adopted across four continents and over 1,000 sites [12]. This is a platform-oriented customer segment, distinct from a buyer seeking a single chest X-ray detection tool. The two vendors may overlap in chest X-ray triage, but their primary documented value propositions address different procurement problems.

Unresolved Questions and Comparison Limits

Several questions cannot be answered from the supplied documentation. Neither vendor provides a direct head-to-head performance comparison on the same dataset. Qure.ai's qXR page cites a publication reporting up to 96 percent sensitivity and 100 percent specificity for missed and mislabeled chest radiography findings, but this is a vendor-cited study, not an independent comparison against Harrison.ai [2]. Harrison.ai's Comprehensive Care page cites a simulation study reporting 32 percent faster report turnaround time for AI-triaged critical cases, but the study design and denominator are not fully specified in the retrieved text [11].

Deployment details also remain partially unresolved. Qure.ai states that qXR can be deployed on cloud or on-premise [1]. Harrison.ai describes deployment of the Harrison Orchestrator to the customer's environment [9]. Neither source provides detailed technical prerequisites for on-premise installation, such as hardware specifications or network requirements. Pricing is not documented in the retrieved materials. Harrison.ai mentions zero platform markup and per-study pricing for HOP [9], while Qure.ai's materials direct buyers to contact the vendor [1]. Any procurement decision would require additional technical and commercial documentation from both vendors.