ESTRO 2024 - Abstract Book

S4415

Physics - Machine learning models and clinical applications

ESTRO 2024

[2] Edmund JM, Nyholm T. A review of substitute CT generation for MRI-only radiation therapy. Radiation Oncology. 2017 Dec;12:1-5.

[3] RaySearch Laboratories, Stockholm, Sweden https://www.raysearchlabs.com/

[4] Zhu JY, Park T, Isola P, Efros AA. Unpaired image-to-image translation using cycle-consistent adversarial networks. InProceedings of the IEEE international conference on computer vision 2017 (pp. 2223-2232).

[5] NONPIMedical AB MICE toolkit v1.1.3 https://micetoolkit.com/

[6] RTsafe, Athens, Greece https://rt-safe.com/

[7] PTW Dosimetry, Frieburg https://www.ptwdosimetry.com/

[8] MVision AI, Helsinki, Finland https://www.mvision.ai/

290

Mini-Oral

Prediction of occult lymph node metastases in oropharyngeal HNSCC

Yoel Pérez Haas 1 , Roman Ludwig 1 , Esmée Lauren Looman 1 , Vincent Grégoire 2 , Laurence Bauwens 2 , Panagiotis Balermpas 1 , Jan Unkelbach 1 1 University Hospital Zurich, Radiation Oncology, Zurich, Switzerland. 2 Centre Léon Bérard, Radiation Oncology, Lyon, France

Purpose/Objective:

Head and neck squamous cell carcinomas (HNSCC) often spread through the lymphatic system of the neck. Treatment does not only include clinically detected lymph node metastases but also lymph node levels (LNL) at risk of harboring occult metastases. The risk of occult metastases depends 1) on the probabilities of the tumor to spread to and between LNLs, and 2) on the ability of biomedical imaging to clinically detect lymph node metastases. Literature values for the sensitivity and specificity of CT, MRI and PET are in the range 70-90%, indicating a false negative rate and a false positive rate around 20% 1 . However, this only represents an average value that is representative for the commonly involved LNL II. Comparing pathological and clinical LNL involvement in oropharyngeal SCC per level reveals that the fraction of occult metastases is higher for the less frequently involved levels III and IV (lower sensitivity). In addition, levels IV and V are clinically positive in 10% of patients or less, which is inconsistent with a sensitivity of 80%. In this work, we develop a statistical model of pathological and clinical lymph node involvement to estimate the risk of occult metastases on a patient-individual level, which addresses the limitations of the traditional concept of sensitivity and specificity.

Material/Methods: