DIAGENS Leads Medical AI Benchmarks and Extends Chest and Abdominal CT Analysis to 509 Imaging Targets

iMed-6.0 brings multimodal reasoning to medical imaging and powers a multi-task CT model, while zero-code tools support specialty AI development.

HONG KONG, Nov. 12, 2026 /PRNewswire/ -- DIAGENS (HKEX: 2526) today announced benchmark results for iMed-6.0, the latest generation of its iMedImage® medical imaging foundation model, and introduced a CT model designed to analyze 509 diseases and imaging findings. In MedBench's multimodal API evaluation, iMed-6.0 Thoth and Aegis took first and second place, respectively, overall and in all three core categories. Thoth also surpassed the highest overall score previously reported on the Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark (MMMU) Test.

The CT model, developed using iMed-6.0 Aegis, applies foundation-model capabilities to hundreds of chest and abdominal CT analysis tasks. Together with the company's zero-code model development tools, it demonstrates how a medical imaging foundation model can support both multi-task analysis and the development of specialty AI.

"We want clinicians and hospitals to build AI around the clinical challenges they know best. With a medical imaging foundation model and zero-code tools, they can train specialty models, adapt them to their workflows, and use them as copilots to support clinical decisions," said Dr. Ning Song, Chairman and CEO of DIAGENS.

Leading results in medical and multidisciplinary evaluation

MedBench is developed by Shanghai AI Laboratory and forms part of the OpenCompass evaluation ecosystem.

In MedBench's multimodal API evaluation, Thoth and Aegis earned overall scores of 64.4 and 61.7, respectively. Both models ranked in the top two for medical visual perception and text extraction, cross-modal semantic understanding and reasoning, and clinical decision support and reasoning. In medical visual perception and text extraction, Thoth and Aegis scored 88.8 and 84.9. These were 20.3% and 15.0% higher, respectively, than the highest score of 73.8 among the other evaluated models.

On the MMMU Test, a multimodal academic benchmark spanning six broad domains and 30 subjects, Thoth achieved 66.2% overall accuracy, exceeding the previously published best result of 65.9%. In the Health & Medicine category, Thoth and Aegis achieved 74.9% and 71.1% accuracy, respectively, placing first and second in the benchmark evaluation.

One model for 509 CT Imaging Targets

The CT model is designed for a defined imaging setting: non-contrast and contrast-enhanced CT examinations of the chest and abdomen. Within this application, it supports automated identification of 509 distinct diseases and imaging findings—263 chest targets and 391 abdominal targets, with overlapping targets counted once.

In an internal validation using nearly 40,000 CT examination–report pairs from multiple hospitals, the model achieved a macro-averaged AUC of 0.855 for zero-shot classification across all 509 targets. For a subset of 249 key targets represented in internationally available public benchmarks, the model achieved a macro-averaged AUC of 0.930.

A repeatable process for building specialty AI

iMed-6.0 processes 3D imaging data, whole-slide pathology images, and temporal imaging data. It supports image interpretation, analysis of anatomy and lesions, and report drafting and review. For supported tasks, outputs can be linked back to specific slices or regions in the source images.

DIAGENS combines these capabilities with iMedLoop™, its medical imaging AI development platform. The platform connects data preparation, model training, evaluation, deployment, and feedback. iMedMaaS® democratizes medical AI by enabling clinicians and hospitals to train, deploy, and scale custom models via a zero-code interface.

As of June 30, 2026, DIAGENS had undertaken 158 model projects and established R&D partnerships with 99 hospitals. Its products and services reached more than 400 healthcare institutions. The company's approach provides clinical teams with a repeatable process for developing models tailored to the imaging tasks they prioritize.

About DIAGENS

DIAGENS (HKEX: 2526) develops medical imaging foundation models, software platforms for specialty AI development, and medical imaging AI products. Its iMedImage model and iMedLoop platform support workflows from imaging data preparation and model training to evaluation, deployment, and feedback. The company works with hospitals and research institutions to develop AI for specific imaging tasks.

diagens.com

SOURCE DIAGENS