DeepHealth and the scan that never happened

The most expensive image in radiology may be the one nobody takes. A scanner sits idle, a specialist cannot be in two places at once, or a patient gets lost between referral and appointment. Even an excellent cancer-detection algorithm has nothing to contribute until the exam exists. DeepHealth, the software business owned by imaging operator RadNet, has made that unglamorous territory part of its proposition.

This is a company for anyone who has wondered why a hospital can possess extraordinary machines and still make a simple appointment feel like a diplomatic negotiation. Its customers are hospitals, health systems and imaging practices. Its users include the person reading your scan, the technologist acquiring it, and the administrator trying to fit everybody into Tuesday.

01 / A startup acquires a second life

The original DeepHealth began in Cambridge in 2015. Co-founder Gregory Sorensen was a neuroradiologist, a former Harvard professor and a former chief executive of Siemens Healthcare North America. Co-founder Bill Lotter brought a Harvard doctorate and machine-learning expertise. The initial focus was mammography, particularly the demanding work of interpreting three-dimensional breast images.

RadNet announced the purchase in March 2020 and completed it in June. Its annual filing recorded $34.6 million in net consideration. The announced deal tied additional share payments to milestones involving regulatory clearance, functionality and productivity. The terms gave technical achievement a commercial consequence.

In November 2023, RadNet launched a broader health informatics portfolio under the DeepHealth name. This explains an otherwise confusing company history: a startup founded in 2015 now supplies the name for a business assembled from several software companies. eRAD contributed imaging information systems; Aidence brought lung AI; Quantib contributed prostate and brain tools. The portfolio subsequently incorporated technologies from Kheiron, iCAD, See-Mode and other acquisitions.

The meaningful change was in the scope of the product. A tool that helps interpret a breast image became part of a plan to connect the people and systems surrounding medical imaging. DeepHealth OS is the cloud-native foundation intended to unify that clinical and operational data.

02 / The algorithm meets the appointment book

Operations Suite deals with the administrative choreography: scheduling, registration, patient messages, billing and analytics. Diagnostic Suite handles image management and the workspace in which clinicians view and interpret studies. Clinical AI adds specialized assistance. Reporting Pro brings findings and measurements into a report that the radiologist reviews and finalizes.

These are different jobs, and the distinction matters. Measuring a suspicious lung nodule requires different expertise from getting an insurance authorization into the right record. DeepHealth’s wager is that buyers want those jobs connected. The company occupies the overlap between enterprise imaging software and clinical AI, where a useful finding must travel through a useful system.

Wichita Radiological Group makes the sales pitch concrete. In November 2025, it selected Operations Suite to replace its legacy radiology information system under a five-year agreement. The group reads more than 500,000 studies a year and operates three outpatient imaging centers. That is an operational purchasing decision, involving daily work across a network, rather than a laboratory demonstration.

“You need a lean, smart system.”Sandra Gonzales, Santa Fe Imaging
Company-published customer testimonial

Reporting Pro reached commercial launch in June 2026, initially in the United States and United Kingdom. It combines speech recognition, clinical AI findings, measurements, draft impressions and quality assurance. DeepHealth describes migration of existing reporting templates and integration with existing PACS and RIS systems. Familiar habits are a product requirement here: a radiologist should spend time examining the patient’s case, rather than transporting numbers between windows.

03 / Expertise can travel farther than people

TechLive addresses another bottleneck: the availability of experienced imaging staff. It supports remote acquisition and collaboration across multiple scanner types and manufacturers. A skilled technologist can help colleagues at another location instead of needing to stand beside every machine. The system received FDA clearance in 2025; DeepHealth announced CE marking and AWS Marketplace availability in February 2026.

That measure is refreshingly literal. A closed room cannot serve a patient. DeepHealth says TechLive contributed to the reduction, which leaves room for other operational changes. The result suggests a practical buying question: where is your imaging capacity being lost, and would access to remote expertise address that particular loss?

Hardware partners offer another route into existing care. GE HealthCare expanded its mammography collaboration with DeepHealth in April 2026. Siemens Healthineers and DeepHealth had announced a letter of intent in December 2024 for ultrasound workflows combining the Acuson Sequoia system and SmartSonography. The attraction is proximity: software arriving where clinicians already acquire images.

04 / A second look has a price

Breast screening remains an instructive example of the clinical business. Safeguard Review uses AI to identify suspicious screening exams that the first interpreting radiologist did not recall, then routes them for additional expert review. The software changes who gets another look. Human judgment remains part of the process.

A July 2026 company account of research published in Radiology describes more than 577,000 screening mammograms, interpreted by 95 radiologists across 109 US sites. Among general radiologists, adjusted cancer detection rose from 3.76 to 4.99 cancers per 1,000 exams. Adjusted recall rates also rose, from 9.06% to 10.40%.

More detection came with more callbacks. A callback means another evaluation for a patient and more work for the service. The positive predictive value of recalls improved as well, according to the study summary, but nobody should read the detection number in isolation. The purchasing question includes what happens after the software raises its hand.

The product range extends beyond breast imaging. Prostate Suite connects MRI analysis to biopsy planning. Neuro Suite measures brain structures and white-matter changes. Thyroid tools assist ultrasound measurement and reporting. The portfolio’s breadth helps explain why DeepHealth speaks to whole imaging organizations, while individual clinicians may encounter only one application.

05 / Buying breadth, then earning the renewal

In March 2026, RadNet acquired Paris-based Gleamer for up to €230 million, including a post-closing milestone. Gleamer brought more than 700 customer contracts in 44 countries and technology spanning routine imaging, including X-ray. This was a substantial purchase of clinical capability and distribution.

The acquired technology keeps evolving under the new name. On September 16, 2026, DeepHealth announced FDA clearance for Chest XRay, an evolution of Gleamer’s ChestView technology. It detects and localizes suspected thoracic findings, including nodules, consolidation and pleural abnormalities. The company said it was commercially available in the US and Europe.

DeepHealth competes with several kinds of supplier. Aidoc offers clinical AI orchestration and workflow integration. Lunit sells breast and chest imaging AI. Established imaging information systems compete for the infrastructure budget. DeepHealth’s distinctive asset is its combination of those categories with a parent that operates the service where the software is used. That relationship provides a testing environment; outside customers still have to find it useful in their own circumstances.

The business earns money through software, AI licenses and related services, with recurring revenue and multi-year agreements. RadNet’s Digital Health segment reported $32.4 million in second-quarter 2026 revenue, including intersegment revenue. Its June-end annual recurring revenue was $105.5 million. ARR is a recurring-revenue measure, not annual revenue, and acquisitions contribute to the changing scale.

The lesson a reader can borrow is an operational one. Pick a bottleneck that people can count: room closure hours, report turnaround, callbacks or unfinished appointments. Put the tool where the work happens. Check the downstream burden as carefully as the immediate improvement. Local product clearance, integration and staffing determine which parts of this approach can travel.

DeepHealth’s most interesting proposition is that radiology improvement requires attention to the whole exam. There is the finding in the image. Then there is everything that allows somebody to find it, review it and act.