ESTRO 2025 - Abstract Book

S94

Invited Speaker

ESTRO 2025

4804

Speaker Abstracts AI in gynaecologic cancers: state of the art and future perspectives Stefania MR Rizzo Imaging Institute of Southern Switzerland, EOC, Lugano, Switzerland. Faculty of Biomedical Sciences, USI, Lugano, Switzerland

Abstract: This presentation provides an overview of how artificial intelligence (AI) is transforming the diagnosis, treatment, and prognosis of gynecologic cancers. It highlights recent advancements, applications, and future directions for AI in this field. AI is currently present in many transversal topics, including imaging acquisitions, radiomics and large language models. When applied to medical imaging acquisitions, many AI-based tools may help to reduce image artifacts, acquisition time, and enhances image quality. Furthermore, in the filed of radiotherapy, deep learning (DL) methods may improve MRI-guided treatment and diagnostic imaging accuracy. Many different AI tools have been applied to gynecologic cancers. For instance, in endometrial cancer AI helps predict tumor grade, lymph node metastases, and recurrence risk using radiomics-based models. In ovarian cancer some AI models have demonstrated the ability to improve risk classification, histological diagnosis, and predict treatment response to chemotherapy. In cervical cancer AI supports lymph node metastasis prediction, staging, and recurrence risk assessment, mostly through radiomics. The current challenges of AI are the availability of large high-quality datasets, regulatory compliance, and seamless integration into clinical practice. Future developments should focus on early detection, risk stratification, personalized treatment, and decision support in oncology. In conclusion, AI is poised to revolutionize gynecologic oncology by enhancing diagnostic precision, treatment planning, and patient outcomes. Continued research, collaboration, and innovation are crucial

4805

Speaker Abstracts How may AI impact our clinical practice in breast cancer screening? Paola Clauser Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria

Abstract:

Most of national breast cancer screening programs in Europe are based on biennial mammography. These examinations are usually evaluated by at least two dedicated radiologists, and women are recalled in case of unclear or suspicious findings. This approach has been proven able to reduce mortality for breast cancer, but has several limitations: the repetitive task of mammography reading leads to fatigue and consecutive errors; in some countries the lack of radiologists led to significant backlogs in screening readings, some breast cancers are difficult to detect on mammography and can be easily overlooked. In addition, the sensitivity of mammography significantly decreases in women with dense breast tissue. Several alternatives have been studies, to overcome this limitations: digital breast tomosynthesis (DBT) slightly improves cancer detection but increases reading times; ultrasound increases cancer detection but also examination times and false positive findings. Magnetic resonance imaging is