On August 5th, the National Internet Information Office released the "Seventh Batch of Deep Synthesis Service Algorithm Registration Information" announcement, announcing that DHC's major model has successfully passed the algorithm registration. This marks a national-level authoritative recognition of its technical strength, innovation capability, as well as the security and compliance of its algorithms.

DHC is a leading enterprise in domestic medical big data and artificial intelligence, focusing on the innovative application of big data and artificial intelligence technologies in the medical field. It provides advanced products and excellent solutions for the entire health care ecosystem. The DHC large model that passed the algorithm registration this time is a multimodal large model aimed at the medical vertical field. It possesses powerful data production capabilities, rapid knowledge iteration capabilities, information mining capabilities, reasoning generation capabilities, and multimodal data analysis capabilities. Centered on AI technology and relying on high-quality multimodal big data, DHC's large model includes text, imaging, pathology, and precision medicine as the four major modal base models. Through deep governance of multimodal data, it empowers applications in multiple scenarios such as assistive decision-making, intelligent research, doctor assistants, health management, and operational analysis.

The DHC large model possesses strong multimodal AI capabilities. For each type of medical data modality, it deeply integrates domain knowledge to form professional AI processing and analysis capabilities and builds base models. It can handle clinical text, multimodal imaging (including CT, MRI, X-RAY, ultrasound, endoscopy, etc.), digital pathology, genetic data (such as NGS, multi-omics), etc. Furthermore, by applying multimodal data and model fusion technology, it constructs the DHC multimodal large model, achieving applications such as multimodal big data platforms, intelligent interpretation of images, and report generation.

DHC Text Large Model

By integrating massive amounts of clinical data and guidelines, consensus, literature, it constructs a 13B-70B language base model for the medical vertical field. It has strong and professional information extraction, understanding and generation, and summarization abilities for clinical cases and text data. This technology has been integrated into DHC's NLP platform and terminology platform, supporting efficient data governance, clinical data structured processing, term standardization, knowledge base construction. At the same time, serving as a clinical research assistant for doctors, it realizes case-assisted writing, specialized disease Q&A, case analysis assisted decision-making, automatic literature summarization, and research hotspot mining.

DHC Imaging Large Model

Using over 2 million multimodal imaging pictures to construct an imaging base model, it supports intelligent analysis of various types of 3D imaging sequences such as CT and MRI, efficiently identifying >150 types of organs, structures, and lesions, covering abdominal, chest, head, pelvic, eye, and other body parts.

DHC Pathology Large Model

Based on 80 million digital pathology patches, it constructs a pathology base with a Transformer architecture, supporting intelligent analysis of more than 30 cancer types, achieving site identification, cancer identification, cancer subtype classification, cancer area segmentation, and nucleus instance segmentation.

DHC Precision Medicine Large Model

It supports automatic identification of entities and relationships in literature, generating clinical annotations, realizing the automatic construction and update of the iCMDB comprehensive knowledge base. iCMDB is a multidimensional knowledge base based on evidence-based medicine levels, having passed ISO13485 medical device management system certification and obtained Singapore HSA ClassA manufacturing license. It has collected more than 2000 diseases, 2000 genes, 3000 drugs, 20000 variants, and 50000 clinical annotations, providing more detailed and reliable evidence-based medical solutions for clinical treatment decisions in molecular oncology, pharmacogenomics, genetics, etc.

Currently, DHC's large model and multimodal AI technology have been widely applied in various medical scenarios including multi-scenario digital diagnosis and treatment, assistive decision-making for precise diagnosis and treatment, comprehensive upgrade of intelligent research, intelligent assistant patient management, etc., providing comprehensive support for the medical field:

 ·  Collaborating with Southern Medical University's Nanfang Hospital to build the first domestic large model-assisted full-hospital multimodal data governance system, achieving efficient management and application of multimodal data, providing comprehensive data support and intelligent assistance for medical decision-making, disease diagnosis, treatment planning, and patient management, significantly improving the quality and efficiency of medical services;

 ·  Supporting Peking University First Hospital in building an acute kidney injury specialty multimodal assisted diagnosis and treatment platform, effectively identifying high-risk groups for acute kidney injury and providing diagnostic and treatment suggestions, significantly improving the efficiency and accuracy of acute kidney injury diagnosis and treatment;

 ·  Relying on the large model-assisted knowledge base support, creating the "Intelligent Report System Platform" at the First Affiliated Hospital of Zhengzhou University, improving the work efficiency of clinical pharmacists while effectively promoting the construction of a precise individualized medication system of "one person, one drug, one dosage";

 ·  Constructing a large model-assisted radiotherapy digital management system for Hainan Provincial Cancer Hospital, achieving unified integration of department business processes and radiotherapy data, significantly improving department management efficiency and promoting clinical and research capabilities;

 ·  Supporting Chongqing University Cancer Hospital in building the first domestic large model-supported intelligent oncology knowledge base, assisting oncology clinicians in clinical diagnosis and treatment decision-making, improving the scientificity and precision of medical decisions, helping researchers quickly and comprehensively access intelligent oncology-related knowledge, promoting scientific discovery and innovation.

In the medical vertical field, the essence of large models lies not in their size but in their specialization and refinement. High-quality real-world multimodal big data and high-performance computing power are the core elements in building domain-specific large models. With the successful algorithm registration, DHC will continue to increase R&D investment, constantly optimizing and upgrading model algorithms to meet the ever-changing healthcare needs. Meanwhile, DHC will also continue to explore the application of large models in more medical scenarios, aiding in the comprehensive improvement of medical service quality and accelerating the progress of medical research. By leveraging digital and intelligent technologies, it continuously promotes the high-quality development of the medical industry.