{"id":18805,"date":"2026-07-20T05:14:26","date_gmt":"2026-07-20T08:14:26","guid":{"rendered":"https:\/\/rtmedical.com.br\/tmp-en-1784535266196\/"},"modified":"2026-07-20T05:14:32","modified_gmt":"2026-07-20T08:14:32","slug":"raidium-ai-oncology-imaging-platform","status":"publish","type":"post","link":"https:\/\/rtmedical.com.br\/en\/raidium-ai-oncology-imaging-platform\/","title":{"rendered":"Raidium brings AI-native oncology imaging to US centers"},"content":{"rendered":"

French startup Raidium has launched Raidium Read (R.Read) in the United States, an AI-native oncology imaging platform that automates the detection, measurement and longitudinal tracking of tumor lesions on CT and MRI studies. Announced on July 16 and aimed initially at cancer research centers, the platform is already live at Moffitt Cancer Center in Florida, one of the leading US oncology institutions, where the company says it replaced a legacy radiomics tool.<\/p>\n

\"Radiologist
AI-native platforms such as R.Read aim to automate the lesion measurements that consume hours of oncology reading time. Photo: Pexels<\/figcaption><\/figure>\n

What R.Read automates in the oncology workflow<\/h2>\n

According to the launch materials, R.Read combines whole-body lesion detection, AI-assisted segmentation with masks the radiologist can edit, and automatic transfer of lesions across follow-up and prior studies. On top of that foundation, the platform produces organ-agnostic automated RECIST measurements across time points and generates structured reports compliant with RECIST 1.1 that can be exported as PDFs.<\/p>\n

Raidium claims the approach cuts inter-reader variability threefold and speeds up lesion annotation by as much as ten times compared with manual workflows \u2014 figures published by the company itself and still awaiting independent validation at scale. Another selling point is frictionless adoption: cancer centers can reportedly start using the tool without lengthy PACS integration projects, according to the announcement.<\/p>\n

Who is Raidium: founders, Curia and a 16 million euro seed<\/h2>\n

Founded in Paris in 2022 by radiologist Paul H\u00e9rent and data scientist Pierre Manceron, both formerly of AI biotech Owkin, Raidium raised a 16 million euro seed round co-led by Newfund and Kurma Partners, with participation from Debiopharm, Founders Future, Galion.exe and the European Innovation Council’s executive agency. The company now operates from Paris and Silicon Valley.<\/p>\n

At the heart of the platform sits Curia, a medical imaging foundation model the company says is trained on a proprietary corpus on the order of one billion real-world images. For oncology lesion segmentation specifically, Raidium reports validation on more than 50,000 annotated lesions, with semi-automatic DICE scores of 0.85 on MRI and 0.74 on CT \u2014 again, numbers published by the vendor.<\/p>\n

Why measuring tumors takes so much radiologist time<\/h2>\n

RECIST 1.1, the international standard for assessing tumor response, requires the radiologist to select target lesions, measure their longest diameters, sum the values and compare the result against the baseline exam and the patient’s historical nadir. At every follow-up, prior studies must be reopened, each lesion re-identified and every measurement repeated consistently \u2014 painstaking work that multiplies in metastatic patients with a dozen or more lesions, and again in clinical trials that rely on blinded independent central review.<\/p>\n

That bottleneck is precisely what foundation models are designed to attack. Rather than training one algorithm per organ or task, these models learn general representations of human anatomy from massive volumes of data and are then adapted to specific tasks such as detection, segmentation and measurement. The logic mirrors that of large language models, which is why the trade press has described Raidium as an attempt to build the “GPT of radiology”. The same modeling family is already reshaping adjacent workflows, from AI auto-contouring software in radiotherapy<\/a> to scheduling tools showing that AI can cut MRI wait times by more than half<\/a>.<\/p>\n

AI-native versus bolt-on AI<\/h2>\n

Most AI tools in clinical use today work as add-ons: the algorithm processes the exam in parallel and pushes results back into a legacy PACS as static DICOM series or notifications. An AI-native platform inverts that relationship \u2014 the viewer is built around the model, and the radiologist interacts with segmentations in real time, correcting contours and propagating measurements across dates. Raidium’s announcement notes that roughly 70% of FDA-cleared AI-enabled medical devices belong to radiology, yet 41% of radiologists feel current tools do not meet their needs, citing a 2025 Philips survey \u2014 a gap the AI-native design is meant to close.<\/p>\n

What it means for radiologists and cancer centers<\/h2>\n

For radiologists, the promise is shifting time from manual measurement to interpretation and clinical decision-making, with measurements that are more reproducible across readers. For research centers, standardized imaging endpoints could lower the cost of clinical trials and make total tumor burden analyses practical, something rarely feasible in routine care. At Moffitt, a staff radiologist said at launch that the unified approach lets the team pursue research projects that once seemed impossible. The launch also lands amid a broader push to widen access to cancer imaging worldwide, exemplified by the Fujifilm and IAEA partnership to expand cancer imaging<\/a> in lower-resource settings.<\/p>\n

Regulatory path and what comes next<\/h2>\n

For now, R.Read is available for research and clinical trial use only. Raidium says it is pursuing FDA 510(k) clearance for a subset of features and expects to announce it before the end of 2026. Until then, the decisive test will be independent validation: performance outside the training distribution, robustness across scanners and protocols, and peer-reviewed publication. Healthy skepticism is warranted \u2014 vendor-reported metrics are a starting point, not a substitute for external evidence, and the competitive landscape of imaging foundation models is getting crowded fast, with academic groups and established trial-reading platforms moving in the same direction.<\/p>\n

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

French startup Raidium launches R.Read in the US: an AI-native oncology imaging platform automating RECIST measurements, already live at Moffitt.<\/p>\n","protected":false},"author":1,"featured_media":18781,"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-18805","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai","category-radiology"],"aioseo_notices":[],"rt_seo":{"title":"","description":"Raidium launches R.Read in the US, an AI-native oncology imaging platform with automated RECIST measurements, already deployed at Moffitt Cancer Center.","canonical":"","og_image":"","robots":"index,follow","schema_type":"Article","include_in_llms":true,"llms_label":"Raidium R.Read: AI-native oncology imaging platform","llms_summary":"French startup Raidium launched R.Read in the US, an AI-native oncology imaging platform built on the Curia foundation model that automates RECIST measurements and longitudinal lesion tracking; live at Moffitt Cancer Center, with 510(k) clearance expected in 2026.","faq_items":[],"video":[],"gtin":"","mpn":"","brand":"","aggregate_rating":[]},"_links":{"self":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/18805\/"}],"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=18805"}],"version-history":[{"count":1,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/18805\/revisions\/"}],"predecessor-version":[{"id":18807,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/18805\/revisions\/18807\/"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/18781\/"}],"wp:attachment":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/?parent=18805"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/categories\/?post=18805"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/tags\/?post=18805"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}