Fraunhofer Institute for Digital Medicine MEVISPress Release
The three-layered structure of the uterus in the MRI after the segmentation process with the CuraMate FADEN work-flow (blue = myometrium, red = junctional zone, yellow = endometrium)
A research consortium led by University Hospital Erlangen and Charité – Universitätsmedizin Berlin has, for the first time, systematically acquired and analyzed MRI scans of the uterus in both healthy and diseased states as part of the RACOON FADEN project. The image data were evaluated using CuraMate, a modular software platform developed by the Fraunhofer Institute for Digital Medicine MEVIS for the annotation, segmentation, and quantification of medical image features.
Worldwide, around ten percent of women of reproductive age suffer from benign changes in the area of the uterus. In the case of adenomyosis, changes in the uterine musculature can cause severe pain and lead to infertility. The RACOON FADEN project aims, for the first time, to develop a method for the early detection of adenomyosis.
Establishing a Biomarker
The focus of the investigation is the three-layered structure of the uterine wall, which consists of the endometrium (uterine lining), a middle layer known as the junctional zone, and a muscular layer; the junctional zone and the muscular layer together form the uterine musculature (myometrium). In the course of adenomyosis, endometrial tissue grows into the muscular layer. The middle, junctional zone is considered a potential biomarker for adenomyosis. “The uterus is a spherical organ shaped like a balloon,” says Prof. Dr. Matthias May, radiologist at University Hospital Erlangen. “Due to its three-dimensional nature, manually determining the thickness of the junctional zone through segmentation is too time-consuming for clinical routine,” he explains.
Automated Annotation Using CuraMate
For this reason, the RACOON FADEN project aims to develop a segmentation workflow for uterine imaging based on deep learning technologies with the help of CuraMate. A web-based software platform developed by the Fraunhofer Institute for Digital Medicine MEVIS constructed as a modular toolkit, CuraMate enables radiologists to annotate nearly all types of medical images. CuraMate incorporates Fraunhofer researchers’ many years of experience in the fields of segmentation and visualization. “We have been continuously developing the toolkit since 2017,” says Dr. Bianca Lassen-Schmidt, project leader for RACOON at Fraunhofer MEVIS. “It includes mature software tools for image annotation, visualization, workflow steps, and review processes, and can be individually configured for each specific use case,” she adds.
Before segmentation, the radiologists must annotate the primary data by tracing the three uterine structures on the MRI images, thus laying the groundwork for training the algorithm. “This form of annotation is very time-consuming,” explains Chiara Tappermann. The doctoral researcher at Fraunhofer MEVIS designed the workflow inRACOON FADEN according to the wishes of the clinicians. The more precisely the annotation is carried out, the better the algorithm can be trained to “trace” the structures independently. Based on this, segmentation can then take place. “With CuraMate, our goal is to automate the segmentation process as far as possible using artificial intelligence,” says Tappermann.
RACOON FADEN brings many new insights
Until now, there has been no systematic investigation using modern technology into what the healthy uterus of a 12- to 30-year-old woman looks like on an MRI scan. “Through RACOON FADEN, we have obtained MRI reference data for the uterus for the first time,” notes Prof. Dr. Sylvia Mechsner, gynecologist at Charité. With the MRI examinations, the radiologists were entering uncharted territory: unlike, for example, a CT scan of the lungs, they did not initially know what they were supposed to segment. For this reason, many different sequences and image types were recorded at the beginning. It became apparent that the uterus can assume very different shapes even during the approximately 40-minute MRI examination and is not positioned uniformly within the body. In response, doctoral researcher Tappermann adapted CuraMate so that the physicians could rotate the three-dimensional image prior to annotation, aligning the axis of the uterus with the image axis. By adjusting the uterine axes across all scans in the study, the segmentations became comparable.
The study provides further new insights into the junctional zone as a biomarker. “This zone, which has always been regarded as having a fixed size in ultrasound, varies in MRI both within a single examination and between the menstrual and ovulation phases,” notes Prof. Dr. May. “The evaluations are not yet complete,” he adds, “but we are seeing a tendency for study participants with severe menstrual pain to show less variability in uterine movement.” RACOON FADEN was also unable to confirm the wall thickness of 12 millimeters for the junctional zone in the early stage of adenomyosis that has so far been reported in the literature.
RACOON FADEN ran from January 1, 2024 to June 30, 2025, and was funded by the Federal Ministry of Education and Research (BMBF, now the BMFTR). In addition to University Hospital Erlangen and Charité – Universitätsmedizin Berlin, 12 other endometriosis centers from the Network University Medicine (NUM) in Germany were involved in the project. The prospective comparative study was conducted using the research infrastructures RACOON (Radiological Cooperative Network) and NUKLEUS (NUM Clinical Epidemiology and Study Platform). In addition to the Fraunhofer Institute for Digital Medicine MEVIS, the German Cancer Research Center (DKFZ), Mint Medical GmbH, and Helmholtz Zentrum München – German Research Center for Environmental Health GmbH also participated in the implementation.