Artificial intelligence in medicine

Central to AI-enabled health research is the consideration of how humans can ultimately make full and proper use of the data generated and processed by artificial intelligence. Visual computing technologies are an essential component of this. For many years now, Fraunhofer IGD has been using machine learning and artificial intelligence methods and procedures to analyze and evaluate vital data, health data and disease-related patient data. These techniques include AI-based medical image processing, visual cohort analysis and anomaly detection in vital data.

Individual health care is a core strategic concern for Fraunhofer IGD. The technologies developed in this context support the transition from classical to personalized medicine by means of visual computing. The specific solutions devised by Fraunhofer IGD cover the entire treatment chain, from diagnosis and therapy right through to aftercare. They involve a combination of imaging methods with data-driven approaches in medicine.

AI for image analysis

Recognizing and marking anatomical structures in medical image data is a prerequisite for the planning and implementation of therapies. 3D image data provide the optimal basis for this, with new learning methods devised by Fraunhofer IGD supporting the medical professionals in the complex process of evaluation and analysis:

• automated segmentation of anatomical structures in image data
• automated detection and marking of anomalies
• informative 3D visualization
• rapid information processing in place of manual analysis
• rapid analysis of image data that would otherwise be difficult to interpret
• applicability to all areas of the human body
• knowledge acquired on a digital basis reduces the number of surgical interventions and of subsequent laboratory examinations

Learn More

Visual cohort analyses

“Cohort” is the term used to describe a group of patients with relevant commonalities. Fraunhofer IGD works in close coordination with clinical users to develop tools for visual-interactive data analysis. These users set the criteria for the patient group. With the help of an individually adapted analysis, it can be seen immediately whether the cohorts are sufficiently meaningful or whether they can be further refined.

Anomaly detection in vital data

By analyzing vital signs such as heart rate, heart rate variability or respiratory rate, the physician can identify sleep quality, stress levels and abnormalities such as sleep apnea or unconsciousness. The extensive collection and continuous evaluation of data facilitates rapid detection of abnormal patterns. This means that medical interventions can be initiated at an earlier stage. Therapy successes can also be better monitored. For reliable detection of abnormalities, we use multiple sensors and situation-dependent algorithms.

Health@Hand is a visual control console for use in care facilities and hospitals. It offers the following functions for the staff:

• digital image of the actual ward
• overview of all health and administrative data of a ward
• visual processing for fast recording and decision-making
• detailed display for individual rooms, individual patients, selected time periods, device overview
• access and interactive exchange via mobile devices possible

Learn More