Chongqing Daily | The first in China! Brain tumor standardized pathology data set released in Chongqing

Reporter Zhang Yizhu

Using AI to assist in the diagnosis of brain tumors, whether the product is reliable or not will be determined by the results of the “examination”. Now the official “examination set” is ready! On April 15, reporters learned that the brain tumor standardized pathology data set independently developed by Jinfeng Laboratory and authoritatively certified by the China Institute for Food and Drug Control was officially released. This is the first standardized pathology data set for brain tumors in China that has been certified by a national statutory inspection agency, and will provide core support for the standardized development of AI products for pathological diagnosis of brain tumors.

▲Jinfeng Laboratory by Fengqi Lake in Western (Chongqing) Science City. (File photo) Photo by chief reporter Long Fan

As AI applications become more and more popular, AI-assisted diagnosis is nothing new. But are the diagnostic results reliable? In this regard, Professor Kong Lingfei, executive director of the Jinfeng Laboratory Smart Pathology Center, said that in the field of medical artificial intelligence, using erroneous, one-sided or incomplete data for large model training will lead to certain deviations in the diagnostic model and output unreliable results, resulting in the inability to meet the demand for accurate diagnosis in clinical applications. Therefore, it is particularly important to collect and process data to establish high-quality data sets.

“Pathological diagnosis is the 'gold standard' for disease diagnosis. High-quality, standardized pathology data sets are the core foundation for medical artificial intelligence technology research and development, product registration and application, clinical transformation and application, and industrial scale upgrading. ”Kong Lingfei said that for a long time, the domestic tissue pathology field has been lacking standardized data sets that comply with national industry standards and have been certified by authoritative organizations, which has become a key bottleneck restricting the innovation and clinical implementation of AI-assisted pathological diagnosis technology.

It is understood that this data set is based on 3,420 pathologically stained full-field digital slices, bringing together 474 typical clinical cases, comprehensively integrating imaging, molecular information and follow-up data to build a multi-modal data matrix； At the same time, a full life cycle closed-loop management system has been established from data collection, pre-processing, accurate annotation to secure storage and authorized access, and is supported by complete technical documentation to achieve standardized management and control of the entire process of identification, quality, and risk. It can be widely used in adult diffuse glioma auxiliary diagnostic product testing, covering core scenarios such as histological and imaging diagnosis, AI molecular prediction, and prognosis assessment.

▲Scientific researchers are busy at the Jinfeng Pathological Precision Diagnosis Center of Jinfeng Laboratory in the Western (Chongqing) Science City. (File photo) Photo by chief reporter Long Fan

The data set released this time can be applied to the Central Inspection Institute to become an "examination set". For example, if a company develops products related to AI-assisted pathological diagnosis, when applying for a medical device license, the China National Medical Examination and Research Institute can use the "examination set" to conduct an "examination" to see whether the product meets the relevant requirements.

It is reported that Jinfeng Lab will also release other data sets as "exercise sets" and make them public, allowing enterprises and scientific research teams to develop and train AI models and improve product performance.

According to industry insiders, the release of this data set marks a milestone step in the construction of high-quality pathology data sets in my country. It not only fills the gap in the domestic industry, but also provides compliant, authoritative and high-quality core data support for the research and development, registration testing, clinical verification and large-scale implementation of AI-assisted diagnosis products for brain tumors in China. It will effectively promote the standardized development of pathological diagnosis AI products and help improve the level of intelligence, precision and homogeneity of disease diagnosis and treatment in my country.