Image-derived AI-based methods to support treatment decision-making for pediatric osteosarcoma patients (ImagAIOS)

Programme

FTI Dissertation

Project coordination

Danube Private University, Ass.-Prof. Dr. Sepideh Hatamikia

Project partners

• Austrian Center for Medical Innovation and Technology (ACMIT)
• Medical University Vienna

Researchers involved at DPU 

• Arezoo Borji, MSc

Abstract

Osteosarcoma (OSC) is the most common malignant bone tumor in children. Beside of the main tumor, approximately 30% of patients have metastases at the time of their diagnosis in lung and further 30% of patients will develop lung metastasis during their life. The actual standard of care for OSC pediatric patients is neoadjuvant chemotherapy, surgery of primary tumor and metastases, and adjuvant chemotherapy. There are several uncertainties concerning the selection of the best and optimal treatment procedures, in particular after complete resection of the primary tumor for patients with metastasis. Therefore, there is an urgent need for technologies that offer more sufficient decision making for a patient-tailored and optimal treatment.

In recent years, with the increasing attention to Artificial Intelligence (AI), several machine learning-based approaches have been developed based on medical imaging data to address different questions for OSC patients. Despite current achievement, several aspects still need to be researched for improving both diagnosis and treatment of such patients.

This project tries to innovate AI methods which can, for the first time, enable the prediction of occurrence of repeated lung metastases, and also tries to improve accuracy of predicting pathologic response to chemotherapy based on the combination of information from various available imaging data of pediatric OSC patients. In this project, innovative AI methods based on both classical machine learning and advanced deep learning techniques will be developed, and a combination of these two approaches will be investigated. Different possible combinations of available imaging data, also in combination with clinical data, will be evaluated in terms of performance.