Researchers from the University of Cologne have developed an innovative digital platform of pathology using artificial intelligence (AI) to analyze lung tissues and cancer diagnostics. The project was headed by a doctor Yuri Tolkach and Professor Rainhard Butnner from the medical faculty and university clinic Cologne.
The developed platform uses algorithms for automatic analysis of sections of tissues of patients with lung cancer. Lung cancer is one of the most common and fatal types of cancer. Nemelcoclet lung cancer (NMRL) is more than 80% of all cases of lung cancer and is considered the second most prevailing and the most deadly type of epithelial cancer.
Successful treatment largely depends on the exact pathological study, in which pathologists analyze biopsy and resection samples. The implementation of AI can significantly optimize this process. Dr. Tolkach notes: “New tools can not only improve the quality of the diagnosis, but also provide new types of information about the patient’s disease, for example, about his reaction to treatment.”
Scientists Teaching AI at the largest available set of high-quality data, which allows technology to quickly analyze biopsy samples, exactly segmenting 11 types of tumor and benign tissues at the pixel level.
Research published in the magazine Cell Reports Medicine demonstrates two key areas of the tool:
- Creation of an accurate model for identifying NMRL types tested and confirmed by patients from several hospitals.
- Identification of four measurable markers in tissue samples that help predict the progression of cancer and patient survival.
In addition, researchers published three datasets to support global lung cancer and develop algorithms.
The team plans to continue validation research in cooperation with five pathological institutions in Germany, Austria, and Japan.
Despite the significant progress in healthcare, it is important to note that the technology is not always accurate in diagnosis. Studies have shown that the accuracy of some commercial tools such as ChatGPT in the diagnosis of diseases in children is only 17%. In addition, popular AI tools, including LLAMA-2-Chat, Vicuna, Medllam2, Bard/Gemini, Claude, and Chatgpt, demonstrated low accuracy in the diagnosis of genetic diseases.