Shuoming Tech Files IPO Prospectus: AI-Powered Diagnostic Platform Achieves 95% Accuracy in Invasive Pulmonary Aspergillosis and Expands into Multi-Disease Applications

Invasive Pulmonary Aspergillosis (IPA) is a dangerous fungal infection that can rapidly disseminate throughout the lungs and other organs, leading to severe complications and death. Clinically, this disease is characterized by “two highs and two lows.”“The Two Highs” refer to the steadily rising incidence rate year by year and the persistently high fatality rate, which pose a serious threat to patients’ life and health. “The Two Lows” represent the current dilemmas in the field of diagnosis: low diagnostic accuracy and low early detection rate.These two major challenges together constitute significant obstacles to the prevention and treatment of invasive pulmonary aspergillosis (IPA).

Based on this, in 2019, the team led by Professors Lv Qingwen and Wang Hua from Southern Medical University initiated research to enhance the diagnosis of IPA. “The traditional diagnostic methods are generally time-consuming and have limited accuracy, which poses a non-negligible risk, especially for patients in critical life-threatening conditions,” said Lv Qingwen.

Invasive Pulmonary Aspergillosis (IPA) progresses rapidly, and any delay in diagnosis can be a matter of life and death. To address this, Lü Qingwen, Wang Hua, and their team are focusing on optimizing diagnostic techniques for IPA by actively integrating artificial intelligence (AI) to improve diagnostic accuracy, thereby securing valuable treatment windows for IPA patients.

Regarding technological innovation, Lv Qingwen stated that the breakthrough of AI in the field of IPA diagnosis lies in the team’s screening and in-depth customization of more than ten leading deep learning AI models worldwide, ultimately resulting in the launch of their independently developed “IPANet” network model.

In terms of accuracy, Lu Qingwen proudly stated: ““By incorporating AI technology, the ‘IPANet’ model has increased the diagnostic accuracy of IPA disease from approximately 80% with conventional methods to 95%.”

In terms of efficiency, the “IPANet” model demonstrates a remarkable speed advantage. Compared with traditional diagnostic methods, which may take several days to complete the diagnostic process, “The “IPANet” model can provide preliminary diagnostic results in just a few seconds.. This leap in speed not only secures a valuable therapeutic window for patients but also provides physicians with more convenient and efficient auxiliary diagnostic tools, significantly alleviating their workload.

It is understood that this model innovatively integrates CT imaging with 15 clinical parameters as comprehensive diagnostic criteria. These parameters encompass multiple aspects, including patient symptoms, physical signs, and laboratory test results, which mutually corroborate and complement the CT findings, thereby establishing a comprehensive and precise diagnostic framework.

Lv Qingwen told VCBeat that, as a pioneer in AI-based diagnosis of invasive pulmonary aspergillosis (IPA), the most pressing challenge was the lack of readily available, high-quality datasets. To address this, the team actively sought collaborations with multiple large tertiary Grade-A hospitals, including Zhujiang Hospital of Southern Medical University. These hospitals not only provided valuable clinical data support but also established close partnerships with the team, jointly advancing the research and development of AI diagnostic technologies for IPA. Through relentless efforts, the team successfully collected and curated nearly 1,000 IPA-related cases, laying a solid foundation for training the “IPANet” model.

During the model optimization process, Lu Qingwen vividly likened the AI model to a “black box.” “Its internal patterns are revealed and its performance enhanced through continuous training and tuning,” said Lu. The team conducted in-depth exploration and iterative refinement across multiple dimensions, including model architecture and parameter configuration, thereby enhancing generalization capability while maintaining model complexity, and ensuring improved computational efficiency without compromising accuracy.

To facilitate the commercialization of research outcomes, Lu Qingwen officially established Guangzhou Shuoming Network Technology Co., Ltd. (hereinafter referred to as “Shuoming Tech”) in October 2023. The company’s core team comprises clinical experts, artificial intelligence (AI) specialists, and professionals dedicated to technology transfer and operations. Notably, although the team is small, with fewer than ten members, it has strategically recruited a seasoned expert in technology commercialization to ensure that research achievements are successfully transformed into market-ready products.

Lv Qingwen told VCBeat that Shuoming Technology’s goal is to develop the IPA diagnostic project into a product with market-launch potential, and on this basis, gradually expand into other medical AI projects:1. An automatic detection and alert system for falls among the elderly, enhance the safety of elderly individuals' daily lives and reduce severe consequences caused by falls;2. Leveraging AI Technology for Predictive Research on Drug Resistance in Lung Cancer；Third, AI technology is used to precisely measure and analyze patients’ symptoms of blepharospasm., providing physicians with more objective and comprehensive diagnostic evidence, thereby facilitating early detection, diagnosis, and treatment of diseases.

From fall detection in the elderly to predicting drug resistance in lung cancer, and further to the analysis of blepharospasm disorders, Shuoming Technology has never ceased its exploration in the field of AI-driven healthcare. Currently, apart from the IPA project, which has entered the productization stage, other projects have completed preliminary research on algorithms and models but have not yet undergone large-scale clinical validation.

In this process, Lü Qingwen, the founder, recognized that the greatest challenge in building a portfolio of medical AI products lies in interdisciplinary communication and collaboration. This requires the team not only to possess dual expertise in clinical medicine and computer science but also to demonstrate strong teamwork and coordination capabilities.

Lu Qingwen cited an example. Initially, the team established reference conditions based on clinicians’ conventional thinking, only to find the results unsatisfactory. Through in-depth analysis, the team recognized that traditional clinical reasoning might have certain limitations in the absence of a clear gold standard. Consequently, the team decided to break from convention and innovatively adjust the data collection and experimental design protocols, ultimately achieving more desirable experimental outcomes.

In the course of collaboration between clinicians and AI experts, although initial differences in thinking styles led to numerous conflicts and challenges, it was precisely these frictions that sparked new inspiration and solutions. Through continuous in-depth communication and intellectual exchange, the team gradually pioneered a new pathway for the efficient and precise diagnosis of IPA disease.

Currently, the AI diagnostic system for IPA is in a critical phase as it moves toward pilot-scale testing.It is reported that the system has successfully completed nearly 20 small-scale preclinical trials, achieving a 100% concordance rate with clinicians’ diagnostic results.However, to achieve the leap from the laboratory to clinical practice, the project team still needs to overcome several significant hurdles.

First, the team must obtain formal approval from relevant government authorities, a prerequisite for ensuring the lawful and safe clinical application of the AI diagnostic system. To this end, Lv Qingwen and his team are intensively preparing the submission materials, striving for full regulatory compliance. Meanwhile, the team is continuously optimizing its algorithms to enhance the system’s stability and reliability, thereby addressing the complexities and variability of real-world clinical environments.

Secondly, the construction and expansion of the database represent another major challenge currently faced. To enhance the generalization capability and accuracy of the AI diagnostic system, the team needs to collect more diverse and representative case data. This not only requires strengthening collaboration with major medical institutions but also leveraging technical means to improve the efficiency of data collection and processing.

In exploring commercialization pathways, Lu Qingwen maintains an open yet prudent stance: he favors securing funding to support subsequent R&D and market promotion, thereby preserving the team’s independence and flexibility. Meanwhile, he also welcomes proposals for technology transfer or collaboration from capable enterprises, aiming to jointly accelerate the rapid development of AI-driven medical technologies.

It is worth noting that, in this process, Shangjun Investment has become an important partner of Lü Qingwen’s team, providing the team with comprehensive support.Chen Yiqun, a partner at Shangjun Investment, told VCBeat that the Shangjun team will not only assist project teams in organizing and preparing application materials in the future, but also leverage its extensive industry resources and experience to provide comprehensive services, including policy interpretation, guidance on application procedures, and communication coordination.. These efforts have undoubtedly provided strong support for the project team to successfully obtain product certification and accelerate the commercialization process.