{"id":17990,"date":"2026-06-01T05:23:59","date_gmt":"2026-06-01T08:23:59","guid":{"rendered":"https:\/\/rtmedical.com.br\/tmp-en-1780302238478\/"},"modified":"2026-06-01T05:29:34","modified_gmt":"2026-06-01T08:29:34","slug":"aium-ai-growing-role-ultrasound","status":"publish","type":"post","link":"https:\/\/rtmedical.com.br\/en\/aium-ai-growing-role-ultrasound\/","title":{"rendered":"AIUM 2026: AI’s Growing Role in Ultrasound"},"content":{"rendered":"

Artificial intelligence in ultrasound is shifting from promise to copilot for whoever holds the transducer. That was the message of the keynote by Alison Noble<\/strong>, of the University of Oxford, at the annual meeting of the AIUM (American Institute of Ultrasound in Medicine) on May 28 in Philadelphia. More than detecting findings in static images, the new generation of AI follows the exam in real time.<\/p>\n

\"Clinician
Video and multimodal AI analyzes ultrasound in real time during the exam.<\/figcaption><\/figure>\n

From image classifier to exam copilot<\/h2>\n

For years, AI in imaging focused on labeling isolated frames. Noble’s point is that the leap now comes from video and multimodal deep-learning methods, capable of automatically analyzing ultrasound video, capturing subtle acoustic patterns, and supporting clinical decisions during both acquisition and interpretation. That changes the nature of the help: instead of a second opinion after the exam, the machine guides the operator while the probe is still on the patient.<\/p>\n

This distinction matters because ultrasound is, among imaging methods, the most operator-dependent. Small variations in angle, gain, or acoustic window change the result. An AI that guides capture targets exactly that Achilles’ heel, bringing the exam closer to a reproducible standard regardless of the operator’s experience.<\/p>\n

The numbers: more accurate gestational dating<\/h2>\n

Noble brought a concrete figure to support the thesis. A study published in 2025 showed that AI support in estimating gestational age reduced the clinician’s mean absolute error from 23.5 to 15.7 days<\/strong>. When the model began offering explanations for its decisions \u2014 so-called explainability \u2014 the error fell further, to 14.3 days<\/strong>. The explanation detail isn’t cosmetic: it indicates that human-machine collaboration yields more when the professional understands the reason behind the suggestion.<\/p>\n

The practical impact is direct. Dating a pregnancy accurately guides decisions on delivery, fetal growth, and interventions. Letting minimally trained professionals perform that dating, without costly equipment, is exactly the kind of democratization ultrasound needs.<\/p>\n

Portable technology for where access is missing<\/h2>\n

One example cited was TraCer, a fully portable system under development in Kenya. It uses a low-cost wireless probe (Konted) paired with a standard Android tablet to capture fetal ultrasound videos. The aim is to bring diagnosis to regions with no sonographers or sophisticated machines \u2014 a mirror of the capacity-building effort for global ultrasound initiatives<\/a> also debated at AIUM 2026.<\/p>\n

That same principle of texture and pattern analysis feeds other fronts, such as the AI model that estimates fetal lung maturity<\/a> from ultrasound. Together, these lines of research sketch a future in which intelligent ultrasound complements \u2014 and sometimes anticipates \u2014 information that once required invasive or high-cost methods.<\/p>\n

Federated learning and human-AI collaboration<\/h2>\n

Noble also highlighted two emerging frontiers. The first is federated learning, in which models are trained on decentralized data from multiple sources without the images ever leaving each institution \u2014 an elegant answer to privacy barriers and the fragmentation of health data. The second is human-AI collaboration, in which the combination of person and algorithm outperforms either one alone. Both ideas point in the same direction: scaling ultrasound AI without forcing hospitals to surrender their data or their clinical authority, two conditions without which adoption rarely survives contact with the real world.<\/p>\n

This framing is consistent with what is seen in other imaging areas: AI doesn’t replace the specialist but raises the quality floor and eases the workload \u2014 a welcome relief in strained systems, as we discussed when covering the chronic shortage of imaging professionals<\/a>.<\/p>\n

Why ultrasound is fertile ground for AI<\/h2>\n

Unlike CT or MRI, ultrasound is cheap, portable, free of ionizing radiation, and produces images in real time \u2014 but it pays for that with heavy operator dependence and acoustic noise. That combination is precisely what makes the method so promising for AI: there’s a real variability problem to solve and a vast volume of exams to train models on. AIUM 2026 devoted several sessions to the theme, including discussions of how AI might even improve the learning of protocols such as eFAST, used in the rapid assessment of trauma.<\/p>\n

Reading in video, rather than in isolated frames, is what brings AI closer to how the sonographer actually works: sweeping, adjusting, and interpreting in motion. The more the algorithm understands the exam’s temporal sequence, the more useful it becomes as a real-time guide.<\/p>\n

What to watch from here<\/h2>\n

For Latin America, where ultrasound is a diagnostic gateway across much of the public and private network, Noble’s agenda is especially relevant. Video AI that guides acquisition can reduce variability between operators, shorten the learning curve, and extend the exam’s reach to remote units. The next steps involve large-scale clinical validation, workflow integration, and clear governance rules \u2014 so the promise of democratization doesn’t run into data bias or a false sense of security.<\/p>\n

Source:<\/strong> AuntMinnie \u2014 “AIUM: AI has evolving role in ultrasound”<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

At AIUM 2026, Alison Noble showed how video AI guides ultrasound in real time and cut gestational dating error from 23.5 to 14.3 days.<\/p>\n","protected":false},"author":1,"featured_media":17986,"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,"fifu_image_url":"","fifu_image_alt":"","footnotes":""},"categories":[102,100],"tags":[],"class_list":{"0":"post-17990","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-ai","8":"category-radiology"},"aioseo_notices":[],"rt_seo":{"title":"","description":"At AIUM 2026, video AI guides ultrasound in real time and cut gestational dating error from 23.5 to 14.3 days. Understand the advance.","canonical":"","og_image":"","robots":"index,follow","schema_type":"Article","include_in_llms":true,"llms_label":"AIUM 2026 AI in ultrasound","llms_summary":"Alison Noble (Oxford) keynote at AIUM 2026: video and multimodal AI analyzes ultrasound in real time; a 2025 study cut gestational dating mean absolute error from 23.5 to 15.7 and then 14.3 days with explainability.","faq_items":[],"video":[],"gtin":"","mpn":"","brand":"","aggregate_rating":[]},"_links":{"self":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/17990\/"}],"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=17990"}],"version-history":[{"count":2,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/17990\/revisions\/"}],"predecessor-version":[{"id":18005,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/17990\/revisions\/18005\/"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/17986\/"}],"wp:attachment":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/?parent=17990"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/categories\/?post=17990"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/tags\/?post=17990"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}