Revolution in cancer diagnosis: KI transforms the pathology!

Revolution in cancer diagnosis: KI transforms the pathology!

Erlangen, Deutschland - As part of an innovative cooperation project between the University Hospital and the Gravina Hospital in Caltagirone, Italy is Integration of artificial intelligence (AI) into the clinical diagnosis of pathology. The aim is to optimize the diagnostic process by using the latest AI algorithms. This development is carried out, while in Germany over 1.4 million cancer are recorded, often as a result of tissue examinations after tumor removal.

Artificial intelligence can support pathologists in identifying cancer more effectively and evaluating tissue samples. However, the use of AI is currently limited in pathology, since many analysis processes are traditionally carried out on the microscope. In order to master these challenges, the Gravina Hospital routinely digitizes all tissue cuts, which significantly improves the availability of digital data.

integration of AI into the workflow

An essential element of the project is the development of a procedure for the automatic integration of the AI ​​analysis into the workflow of the pathological laboratory. The tissue samples are processed into thin cuts and then digitized so that the diagnosis is made on the computer monitor. As soon as new scans enter into the laboratory information system (LIS), the AI ​​analysis is automatically activated. Pathologists can also request "on-demand" analyzes, which improves flexibility and efficiency in the diagnostic process.

The results of the AI ​​analysis are visualized in the form of heat maps in the LIS in order to emphasize cancer-like regions. The project aims to improve the accuracy of the algorithms and to promote the integration of deep learning models in other pathology departments.

technological innovations in the diagnostic process

technological innovations have significantly accelerated the diagnosis of cancer, in particular through the digitization of histopathology cutting and the use of deep learning. According to an article on pmc.ncbi.nih.gov , the traditional diagnostic pathology is lengthy and error -prone, which in view of the increasing cancer and stress creates an urgent need for action.

deep learning models are suitable for clinical tasks that require high reproducibility and low fault tolerance. Historical developments in the area of ​​deep learning, from the first neuronal networks to modern convolutional neural networks (CNNS), have significantly improved diagnostic accuracy. For example, a developed model for grading assessment in prostate cancer achieved an accuracy of 0.7, while human pathologists could only record an accuracy of 0.61

In addition, graphic -based neural networks (GNNS) offer promising approaches to improve both the performance and interpretability in digital pathology. These networks model relationships between objects and have already achieved success in the field of cancer diagnosis, including in the evaluation of colorectal cancer with an accuracy of 97%.

The role of big data and AI in medicine

The digital change leads to an explosion of available data that can be used effectively in medicine. Fraunhofer IKS emphasizes that artificial intelligence can quickly analyze large amounts of data, which leads to an individualization of therapies and early diagnoses. This development is supported by the networking of medical and non-medical data, which enables efficient and rational decisions.