{"id":19340,"date":"2026-09-14T05:20:35","date_gmt":"2026-09-14T08:20:35","guid":{"rendered":"https:\/\/rtmedical.com.br\/tmp-en-1789374034941\/"},"modified":"2026-09-14T05:20:41","modified_gmt":"2026-09-14T08:20:41","slug":"epsilon-health-ai-native-radiology","status":"publish","type":"post","link":"https:\/\/rtmedical.com.br\/en\/epsilon-health-ai-native-radiology\/","title":{"rendered":"Epsilon Raises $27.6M for AI-Native Radiology"},"content":{"rendered":"

A San Francisco startup has come out of stealth announcing it raised nearly $28 million<\/strong> to attack the US radiologist shortage along an unusual path: instead of selling algorithms to practices, Epsilon Health<\/strong> wants to be the practice. The Series A, disclosed on September 10, 2026, was led by AlleyCorp and marks the sector’s second significant bet in under two weeks on what the market has started calling “AI-native” radiology<\/strong>.<\/p>\n

\"Radiologist
The reading workstation is the bottleneck AI-native startups promise to unlock. Photo: Tima Miroshnichenko\/Pexels<\/figcaption><\/figure>\n

Who Epsilon is and who joined the round<\/h2>\n

Founded in 2024, the company is run by Rustin Rassoli<\/strong>, who studied mathematics and computer science at the University of Texas at Austin and grew up inside his father’s imaging practice. “When a diagnosis is missed or delayed, it can quickly change the course of someone’s life,” he said in the announcement, explaining his choice of problem. His read of the US market is blunt: the radiologist shortage “is putting real lives at even greater risk \u2014 the system is collapsing, and patients are bearing the brunt of it.”<\/p>\n

Alongside New York-based AlleyCorp, which led the round, participants included Uncork Capital, Renegade Partners and SemperVirens. Jack Altman<\/strong>, brother of the OpenAI founder, also joined while serving as managing partner at Alt Capital. The cap table says something about the thesis: this is technology money going into what is, ultimately, a medical service provider.<\/p>\n

The bet: be the practice, not just the algorithm<\/h2>\n

Epsilon says it differs from other radiology AI companies, which build models for a single disease or imaging modality. Its pitch is “purpose-built AI”<\/strong> integrated directly into the work of board-certified radiologists, able to accelerate interpretations while “preserving how providers already deliver care.” Put differently: the product is not the detector, it is the whole reading service, with the algorithm underneath.<\/p>\n

The market argument the company leans on is familiar and strong: roughly 75% of all medical decisions<\/strong> rely on radiology to diagnose and treat disease. If exam demand grows and the number of physicians reading them does not keep pace, the bottleneck becomes delayed treatment. Selling reading capacity rather than a software license is the most direct way to capture the value of clearing that bottleneck.<\/p>\n

Why the standalone-algorithm model stalled<\/h2>\n

There is an economic reason for the pivot. An algorithm that detects one specific condition has to be bought by someone, integrated into someone’s PACS and \u2014 the hardest step \u2014 paid for by someone. While AI reimbursement advances slowly, the developer stays trapped in long sales cycles and a benefit that shows up in the customer’s workflow rather than its own. When the company is the provider, the productivity gain converts to margin directly.<\/p>\n

It is not an isolated move. Days earlier, Australia’s Harrison.ai surfaced backing a US teleradiology group in an arrangement that drew debate over conflict of interest<\/a> precisely because it mixes algorithm developer and clinical service provider. Worth noting: US rules require physician ownership of entities providing clinical services, which forces these structures to separate formally who owns the practice from who supplies the technology. On the infrastructure side capital has been flowing too, with Scan.com raising $220 million<\/a> to build a scheduling and imaging network.<\/p>\n

The counterpoint: how much AI actually delivers<\/h2>\n

Two caveats deserve room. The first concerns performance. A speed promise is easy to announce and hard to sustain in production, where case mix, acquisition quality and population profile change everything. Recent assessments show a gap between expectation and result \u2014 that is what radiologists reported when measuring what mammography AI actually delivered<\/a>.<\/p>\n

The second concerns the shortage diagnosis itself. It is real but not uniform: it concentrates in specific subspecialties, shifts and regions, and it coexists with signs of a hot market, such as the 6.6% rise in radiologist pay<\/a> in the US. A company promising to solve “the shortage” with throughput has to show, in practice, that the gain comes from a better workflow and not merely from shorter turnaround at the cost of shallower review.<\/p>\n

What it means beyond the US market<\/h2>\n

The arrangement Epsilon proposes already exists elsewhere in other clothing. In Brazil, teleradiology<\/strong> companies have operated for years as providers that combine their own clinical staff, an exam distribution platform and, more recently, triage algorithms. The American difference is the scale of the capital and the fact that AI enters as a design premise rather than a module bolted on afterward.<\/p>\n

For a department leader, the practical question is not whether to buy AI but where it enters the workflow: queue prioritization, pre-populating a structured report, retrospective quality control or autonomous detection. Each of those positions carries a different regulatory requirement and a different clinical risk. And there is a governance lesson in the AI-native model that holds regardless of vendor: when whoever runs the algorithm also signs the report, performance auditing has to be external, continuous and documented \u2014 otherwise there is no way to distinguish a genuine productivity gain from simply spending less time looking at the image.<\/p>\n

Source:<\/strong> AuntMinnie<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

Epsilon Health raised nearly $28M to be the radiology practice itself, not just the algorithm. Inside the AI-native model and its limits.<\/p>\n","protected":false},"author":1,"featured_media":19308,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"ngg_post_thumbnail":0,"_rt_cluster":"","fifu_image_url":"","fifu_image_alt":"","footnotes":""},"categories":[102,100],"tags":[],"class_list":["post-19340","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai","category-radiology"],"aioseo_notices":[],"rt_seo":{"title":"","description":"Epsilon Health raised nearly $28M to be the radiology practice itself, not just the algorithm. Inside the AI-native model and its limits.","canonical":"","og_image":"","robots":"index,follow","schema_type":"Article","include_in_llms":true,"llms_label":"Epsilon Health raises $27.6M for AI-native radiology","llms_summary":"Startup Epsilon Health raised nearly $28 million in a Series A led by AlleyCorp to operate as a radiology practice with integrated AI rather than selling algorithms.","faq_items":[],"video":[],"gtin":"","mpn":"","brand":"","aggregate_rating":[]},"_links":{"self":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/19340\/"}],"collection":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/"}],"about":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/types\/post\/"}],"author":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/users\/1\/"}],"replies":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/comments\/?post=19340"}],"version-history":[{"count":1,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/19340\/revisions\/"}],"predecessor-version":[{"id":19342,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/19340\/revisions\/19342\/"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/19308\/"}],"wp:attachment":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/?parent=19340"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/categories\/?post=19340"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/tags\/?post=19340"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}