Stereotactic radiotherapy (SRT) is the current standard of care for brain oligometastases. SRT is characterized by a high dose delivered in less than five fractions and with a steep dose gradient to surrounding healthy brain tissue. However, follow-up MRI is often ambiguous due to the frequent presence of treatment-related changes such as radiation necrosis or pseudoprogression. Importantly, a valid assessment of local control is essential for the indication of further systemic therapy such as modern targeted therapy or immunotherapy.

The practice of medicine is changing with the development of new Artificial Intelligence methods of machine learning. Nowadays, radiology, as well as radiotherapy, strongly relies on the help of computers, especially when dealing with large 3D image data. Finding important image content that is often hidden to the naked eye is called radiomics, which means extraction of image or object features (descriptors) that quantitatively characterize the image or its parts in mathematical terms.

It is reasonably assumed that there will be a significant increase of incidence of patients with brain metastases indicated to SRT. Exact and valid evaluation of follow-up MRI after targeted SRT represents an unmet clinical need which is increasingly important for the individual patient in respect to increasing possibilities of modern systemic treatment of disseminated disease. Moreover, due to the limited availability of advanced MRI or PET/CT methods for brain imaging for further imaging of equivocal findings on standard contrast enhanced structural MRI, the development of other approaches is recommended.

The aim of presented study is to develop a radiomic model for the evaluation of contrast enhanced structural brain MRI after SRT of brain metastases in terms of differential diagnosis of radiotherapy-related changes and true progression in order to increase sensitivity and specificity of MRI evaluation. Specifically, a retrospective cohort of patients treated by stereotactic radiotherapy for brain metastases will be created for subsequent development of radiomic model. Prospective cohort with regular standardized follow-up will serve for validation.