How AI is Transforming Medical Diagnostics at Herne’s St. Elisabeth Hospitals

The St. Elisabeth Group hospitals in Herne are integrating artificial intelligence (AI) across radiology, anesthesiology, and gastroenterology to reduce diagnostic times and improve patient safety. According to medical directors at the group, these AI tools function as decision-support systems that optimize workflows and automate measurements, though they do not replace the final clinical judgment of physicians.

AI Integration in Radiology and Brain Imaging

At Marien Hospital Herne, AI is now embedded within CT and MRI hardware to increase image quality and accelerate scan speeds. Prof. Dr. Lars Schimmöller, Director of the Institute for Diagnostic, Interventional Radiology and Nuclear Medicine, states that these advancements have reduced MRI examination times from up to 60 minutes in the past to between 10 and 30 minutes today.

Specific AI applications currently in use include:

• Lung CTs: Automated detection of abnormalities for physician review.
• Cardiac Diagnostics: Automatic mapping of heart contours, replacing previous manual tracing.
• Prostate MRI: Automated volume calculation to help distinguish between benign enlargement and potential prostate cancer.
• Neurological Imaging: Identification of brain regions suffering nerve cell loss to assist in diagnosing degenerative diseases like Alzheimer’s.

Hospital staff maintain a “human-in-the-loop” approach. Doctors and radiological assistants review and adjust all AI-generated suggestions to ensure accuracy.

Predicting Hypotension in Surgical Settings

Since 2019, the Marien Hospital Herne has utilized AI in operating rooms to monitor blood pressure data during anesthesia. Prof. Dr. Ulrich Frey, Director of the Center for Anesthesiology and Operative Intensive Care Medicine, reports that the system predicts impending drops in blood pressure and identifies the likely cause.

This capability is critical during high-risk procedures, such as aortic surgeries, where rapid blood pressure drops can lead to immediate complications. According to a study conducted by the hospital, the use of this AI technology resulted in five times fewer blood pressure drops during operations.

Gastroenterology and Internal Medicine Automation

The St. Anna Hospital Herne has deployed AI across several specialized departments to increase detection rates of early-stage diseases:

Colorectal Screening: In all four examination rooms, AI is used during colonoscopies to mark potential polyps. Dr. Viktor Rempel, Chief Physician of the Clinic for Gastroenterology, notes that the AI is specifically trained to find very small polyps that are difficult for the human eye to spot. While experienced doctors use it for verification, the tool is particularly valuable for physicians in training.

Cardiac Ultrasound: The Clinic for Internal Medicine uses AI to automatically calculate the heart’s pumping capacity. Dr. Panagiota Zgoura, Chief Physician, explains that this removes the need for manual calculations, saving time for the patient. This system has been in routine use for two years.

Research and Early Diagnosis in Rheumatology

The Rheumatology Center Ruhrgebiet is currently using AI within a research framework to accelerate arthritis diagnosis. The center employs two primary technologies:

• Ultrasound Robotics: A robot performs the ultrasound and suggests a classification of findings, which a physician must then confirm.
• Thermography: AI analyzes 3D images of skin temperature. Because arthritis typically causes joint swelling and overheating, these thermal differences allow for faster diagnosis of inflammatory rheumatic joint diseases.

Prof. Xenofon Baraliakos, Medical Director of the Rheumatology Center, emphasizes that these tools are designed to make diagnoses faster and more accurate, not to replace the physician.

Medical AI Guardrails and Limitations

Physicians within the St. Elisabeth Group maintain that “self-learning” models are not permitted in a clinical setting due to safety and regulatory requirements. Instead, the group uses models trained on anonymized patient data—such as blood pressure parameters and laboratory values—before they are deployed in the clinic. This ensures the AI operates on a fixed, validated logic rather than evolving unpredictably during patient care.