CASC Securities: Heavy cooperation in AI medical sector frequently appears to deepening the application scenarios.

CSC Securities issued a research report stating that the commercialization of AI+ healthcare both domestically and internationally is accelerating, with key breakthroughs in various tracks landing. Domestically, the focus is on AI healthcare large models, AI medical imaging, AI pharmaceuticals, and smart wearables. AI-assisted diagnosis can improve diagnosis efficiency, consistency, and assist in primary healthcare, demonstrating clear value in reducing costs and increasing efficiency. AI medical imaging is moving towards integrated management of multiple disease types and the entire process; AI-driven drug development (AIDD) has entered the clinical verification stage from early discovery, requiring attention to platform technology validation, deep cooperation models with top pharmaceutical companies, and real improvement data in clinical stage success rates/speed. Key points from CSC Securities are as follows: Overseas, the highlight this month is the acceleration of the commercialization of AI+ healthcare from laboratory verification to real-world clinical landing and full chain commercialization, with innovation and standardization developing in parallel, and key breakthroughs in multiple tracks landing. Clinical landing and standardization of AI medical imaging and equipment are accelerating. Google, in collaboration with the NHS, released a real-world study on AI for breast cancer screening, with AI able to identify 25% interval cancers and reduce the workload of radiologists by about 40%; Stanford Merlin's 3DCT model has an accuracy rate of over 81% in diagnosis and extends to predicting risks of chronic diseases; Butterfly Network's AI gestational age estimation ultrasound tool received FDA approval, lowering the threshold for primary care operations; FDA updated the AI medical device list, further clarifying the safety framework in multiple scenarios. Breakthroughs in AI clinical decision-making and diagnostic assistance have been achieved. NIH-funded Harvard's multi-modal AI tool for predicting domestic violence risks had an accuracy rate of 88% three years in advance embedded in EMR. Continued escalation in AI drug development infrastructure and cooperation. Roche deployed over 2,100 NVIDIA Blackwell GPU to build a hybrid AI infrastructure; Lilly expanded its collaboration with Insilico Medicine to $2.75 billion, advancing AI-driven oral drug innovation development. Domestically, this month's main focus is on AI healthcare large models, AI medical imaging, AI pharmaceuticals, and smart wearables. In the field of AI healthcare large models, the first large model for operation management in a public hospital nationwide landed at Anzhen Hospital, with management report generation reduced from several days to 5 minutes; JD HEALTH's "Jingyi Qian Inquiry" once again topped the MedBench evaluation, continuing to lead in multi-modal capabilities. In the field of AI medical imaging and diagnostic assistance, Alibaba's Damo Academy collaborated with Neusoft Medical to embed a multi-cancer early screening algorithm in CT equipment, promoting universal access at the grassroots level; "The first stock in medical imaging large models" De Shi Life Science listed on the Hong Kong Stock Exchange, achieving 100% sensitivity and specificity in AI-based karyotype analysis for detecting abnormal numbers. In the field of AI pharmaceuticals, Lilly's $2.75 billion bet on INSILICO set a new high for domestic AI pharmaceuticals' external authorization; INSILICO has also expanded its cooperation in the central nervous system field, entering a key period of commercial verification for the industry. In the field of smart wearables and life monitoring, Zumu released the AI health ring Glow, pioneering fingertip ECG monitoring function; Andon's 5X can provide hours to days of early warning for emergencies such as heart attacks, moving the industry from passive recording to active warning. In terms of policy and infrastructure, Shanghai has included medical AI in its core infrastructure agenda; multiple departments in Jiangsu are working together to accelerate the landing of "AI+ healthcare"; the National Health Commission has made it clear for the first time that AI-assisted papers must be declared; and the government work report has listed brain-machine interface in the list of industries to nurture in the future. In terms of industry consensus, the world's first AI hospital consensus was released, specifying five core characteristics of AI hospitals. Currently, medical companies at home and abroad continue to deploy AI products and services, including medical imaging, clinical decision support, precision medicine, health management, medical informatization, drug development, and medical Siasun Robot & Automation, empowering all aspects of the pharmaceutical and medical industries through cost reduction, improved patient experience, and reduced disease risk. AI is transitioning from a role of "technical assistance" to being the core driver of "value reshaping" and "efficiency revolution" in the medical industry, with its commercial value permeating from research and development to clinical, payment, and patient sides comprehensively. The core theme of AI healthcare can mainly be seen in the following aspects: In terms of scenario demand, the investment logic of AI healthcare mainly lies in addressing industry pain points. AI-assisted diagnosis can improve diagnosis efficiency, consistency, and assist in primary healthcare, demonstrating clear value in reducing costs and increasing efficiency; in fields such as early screening for cancer, companion diagnostics, and treatment prediction, it can help achieve precision medicine; in the field of new drug development, it is expected to bring significant returns to pharmaceutical companies. In terms of technical evolution and commercialization maturity, the development maturity of AI varies in different application scenarios. AI medical imaging has evolved from assisting in the diagnosis of a single disease (such as lung nodules) to integrated management of multiple diseases (such as full disease triage for abdominal CT) and the entire process (screening, diagnosis, treatment prediction), requiring attention to the expansion capability of the product matrix, the depth of integration into clinical workflows, and the potential for upgrading to treatment decision support; AI-assisted drug development (AIDD) has moved from early discovery to clinical verification, necessitating attention to platform technology validation, deep cooperation models with top pharmaceutical companies, and real improvement data in clinical stage success rates/speed. Medical large models and multi-modal AI, such as models released by Google and domestic manufacturers, can handle multidimensional medical data such as text, images, and voice, requiring attention to the accuracy of the models in professional fields, the integration capabilities with existing hospital information systems (HIS/PACS), and the potential to build an ecosystem as a "medical intelligence entity." Recommendations for focus: 1) AI Pharmaceuticals: XTALPI, PharmaResources, Hitgen Inc., INSILICO, etc; 2) AI medical imaging and diagnostic assistance: Shanghai United Imaging Healthcare, Beijing Wandong Medical Technology, etc; 3) Medical informationization and smart hospitals: Goodwill E-Health Info, B-Soft Co., Ltd., DHC Software, Winning Health Technology Group, etc; 4) Internet medical and health platforms: JD HEALTH, ALI HEALTH, PA GOODDOCTOR, etc; 5) Precision medicine and AI-driven medical services: Guangzhou Kingmed Diagnostics Group, Shanghai Runda Medical Technology, Meinian Onehealth Healthcare Holdings, etc; 6) Technology/data platform enterprises: iFlytek Medical, YIDU TECH, etc. Risk warning: the risk of the landing pace of large AI model technology in the software and service fields being lower than expected; the risk of the slower pace of policy implementation related to AI healthcare; the risk of the development of the AI healthcare industry falling short of expectations; the risk of intensified industry competition; the risk of delays in R&D and production progress due to restrictions imposed by external suppliers on hardware or software; the risk of changes in macroeconomic conditions.