已获备案通知书。 本文为IPO早知道原创 作者｜ Stone Jin 微信公众号｜ipozaozhidao 据IPO早知道消息，杭州德适生物科技股份有限公司（以下简称"德适生物"）于2026年1月6日更新 招股书，继续推进港交所主板上市进程，华泰国际担任独家保荐人。 这意味着，德适生物继续向"医学影像大模型第一股"发起冲击——1月5日，中国证监会国际合作司 已披露了《关于杭州德适生物科技股份有限公司境外发行上市及境内未上市股份"全流通"备案通知 书》， 即德适生物已取得港交所聆讯的前置条件。 根据最新版本的招股书， 2025年前三季度，德适生物的营收为1.12亿元，较2024年同期的0.20亿 元增长约470%，同期的毛利率亦进一步提升至75.9%。 而之所以德适生物的业绩能实现如此大幅度的增长，一定程度上还是得益于其在所处赛道的稀缺性， 以及政策持续利好的加持。毕竟，医疗是当下大模型落地价值最大的行业之一，同时其又具有相当之 高的准入门槛。 成立于2016年的德适生物专注于医学影像AI与智能装备领域，现已开发出多元化的产品组合，包括 六款医学影像软件产品、三款商业化医疗设备、四款主要试剂及耗材以及两款技术许 ...

1月5日，中国证监会国际合作司发布《关于杭州德适生物科技股份有限公司境外发行上市及境内未上市 股份"全流通"备案通知书》。德适生物拟发行不超过3100.4万股境外上市普通股并在港交所上市。该公 司27名股东拟将所持合计8088万股境内未上市股份转为境外上市股份，并在港交所上市流通。 招股书披露，德适生物一直专注医学影像AI行业。公司专有的iMedImage™具有跨模态预训练架构，使 单一统一模型能支持19种医学影像模态(如染色体、CT、MRI、超声及病理)，并覆盖超过90%的临床医 学影像场景，包括生殖健康、血液系统恶性肿瘤及放射卫生。 根据弗若斯特沙利文资料，德适生物自主研发的iMedImage™是世界上参数规模最大的通用型医学影像 基座模型，已成功实现商业化，而且被认定为世界上首个商业化跨模态医学影像基座模型。 业绩方面，德适生物连续亏损。2023年、2024年及2025年前三季度，德适生物收入分别为5284.4万元、 7035.2万元、11161.6万元，年/期内亏损分别为5611.6万元、4337.5万元、3664.9万元。尽管公司营收增 长，但净利润连续亏损，两年多累计亏损1.36亿元。 从收入构成 ...

Core Insights - The article highlights the launch of iMedImage™, a large-scale medical imaging model developed by Deshi Bio, which is based on 80 million data points and supports 19 modalities, covering over 90% of clinical scenarios [1] - iMedImage™ significantly reduces data dependency and development costs by over 90%, requiring only hundreds of images and a few days for training specialized models compared to traditional methods [1] - Deshi Bio has received the Sullivan Global Medical Imaging Foundation's Innovation Award for iMedImage™ and aims to lower the barriers for AI applications in medical research and clinical settings [1] - During the event, Deshi Bio signed strategic cooperation agreements with Wutongshu, Inspur Group, and ZTE Corporation to advance the research and industrial application of AI in medical imaging [1] Summary by Categories Product Development - iMedImage™ is the largest general medical imaging model globally, trained on 80 million data points [1] - The model supports 19 imaging modalities, including chromosome, CT, and MRI, and covers over 90% of clinical scenarios [1] - The training process is automated and requires minimal data input, significantly reducing the time and resources needed for model development [1] Cost Efficiency - iMedImage™ reduces the data dependency for specialized models to just hundreds of images and training time to a few days [1] - Development costs are decreased by over 90% compared to traditional specialized models [1] Industry Recognition and Collaboration - The model has won the Sullivan Global Medical Imaging Foundation's Innovation Award, indicating its significance in the field [1] - Deshi Bio has established strategic partnerships with Wutongshu, Inspur Group, and ZTE Corporation to promote the application of AI in medical imaging [1]