The EU-funded project EuCanImage, in which the EACR is a partner, just launched a large-scale study to perform a direct analysis of the impact of the annotator differences on AI training. EuCanImage seeks radiologists to support this groundbreaking collective segmentation initiative.
This project entails, among other tasks, the anotation and segmentation of cancer and noncancer lesions from multiple imaging examinations (including liver CT, liver and pelvic MRI, mammography, and breast MRI) of patients with the following diseases:
- Hepatocellular carcinoma,
- Rectal cancer, and
- Breast cancer.
The EuCanImage repository contains more than 20,000 imaging examinations from multiple European partners. The consequences of inter- and intra-observer variability in data annotations is a known factor in achieving a consistent and robust AI model in medical imaging. Although there are a few studies that explore inter- and intra-rater differences, they do not analyze the effect of multiple annotators in AI training but mostly focus on showing how much these annotators differ from each other.
The annotators taking part in the EuCanImage study would be enrolled as additional, temporary members of the University of Pisa (UNIPI) team, and would be granted secure personal access to the EuCanImage online image processing platform, remotely accessible from any computer with a reasonably high-speed Internet connection.
The collective segmentation will be performed free of charge, and will be granted co-authorship of at least one paper on the topic published by the EuCanImage consortium in international journals. We expect that each researcher will have to annotate 100 mammograms or 30 liver CTs (up to 3 lesions) over a time frame of several months.
If you are interested in joining this ambitious project or would like to learn more about this initiative, please reach out to Roberto Francischello and Lorenzo Faggioni.
Click here to visit the call on the EuCanImage website
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952103
