The early and precise diagnosis of endometriosis is crucial for women's health and quality of life. Researchers at Dortmund University of Applied Sciences and Arts are using hyperspectral imaging and AI methods to improve medical care for those affected.
Endometriosis is a chronic disease in which tissue that resembles the lining of the uterus grows outside the uterus. Affected women often suffer from severe pain, but the symptoms vary greatly. This is why it sometimes takes a long time for the disease to be diagnosed by a doctor. With consequences: If the tissue is not removed, it can even lead to infertility. According to the Robert Koch Institute (RKI), 10 to 15 percent of women suffer from endometriosis.
Minimally invasive surgery using endoscopy is considered the gold standard for treatment. However, until now, only the visual endoscope image has been used to decide whether and where exactly an endometriosis lesion is present. This carries risks. "We also rely on hyperspectral image analysis for tissue classification," says Stefan Patzke, research associate in the "HSI4MIC" project at the Faculty of Information Technology at Fachhochschule Dortmund.
255 instead of three spectral bands
The hyperspectral camera used captures up to 255 spectral bands. The purely visual endoscope image only has three bands (red, blue and green). In addition to the significantly improved spectral resolution, there are also several other spectral bands in the near-infrared to UV range that are not even perceptible to the human eye. In this "spectral fingerprint", Stefan Patzke uses artificial intelligence to search for characteristic features of an endometriosis lesion. "Our goal is to provide assistance for doctors," emphasizes Stefan Patzke. In future, new sensor technology and the corresponding software will be used to provide information on affected tissue directly during the endoscopic examination. "The doctors can then check these areas again in detail." The aim is to minimize endometriosis residues in the body and thus reduce the number of follow-up operations.
The clinics also have a great interest in this. Fachhochschule Dortmund is cooperating in this project with hospitals in the region - the Klinikum Dortmund and the Endometriosis Center at Marienkrankenhaus in Schwerte. They have been equipped with a hyperspectral camera and provide the high-resolution spectral images of the tissue that Stefan Patzke is working with. The corresponding results should be available next year. "I assume that our approach will make it possible to reliably identify the damaged tissue," he says confidently. The next step is the technical integration of the hyperspectral camera into the endoscopic tool. "This will be easier if we know exactly which spectral bands are relevant for detection."
The relevance of his research for everyday medical practice was also acknowledged by the award jury at the Fachhochschule Dortmund's DART Symposium(Opens in a new tab) . Stefan Patzke was presented with the Award for Young Researcher 2024.
Background
The "HSI4MIC" project (hyperspectral image analysis for tissue classification in minimally invasive surgery) is funded by the Federal Ministry of Education and Research. The project is headed by Prof. Dr. Jörg Thiem, Vice-Rector for Research and Transfer at Fachhochschule Dortmund. Stefan Patzke is supported in the project by a student assistant. In addition to the Dortmund Clinic and the Endometriosis Center of the Marienkrankenhaus in Schwerte, the medical technology company C.R.S. iiMotion GmbH is also involved as a partner. The project will also lead to further student work in the field of hyperspectral imaging. "We stand for the unity of research and teaching," emphasizes Jörg Thiem, who teaches in the Faculty of Information Technology. This is why research takes place visibly and with the involvement of students.