Multi-Omics Integration for Personalized Radiotherapy
Accelerating Progress: Fast Deployment of AI-Based Studies
Personalized Radiotherapy (RT) through artificial intelligence (AI) presents promising opportunities for enhancing tumour control probability and minimizing normal tissue toxicity. However, the construction of robust AI models necessitates significant data, often derived from multiple centers. The absence of standardized data, inconsistencies in naming conventions, and variations in acquisition parameters can lead to challenges such as missing, mislabeled, or corrupted data. To overcome these obstacles, we leverage state-of-the-art AI solutions, utilizing advanced algorithms and methodologies to ensure precise curation and preparation. This includes the application of complex analysis workflows, automated integrity checks, and content-based data classifications. By implementing these cutting-edge techniques, we achieve a seamless integration of various data sources, thereby laying the groundwork for fast deployment of AI studies in personalized RT.
The figure illustrates our approach to personalized radiotherapy, integrating radiomics, dosiomics, and biomolecular omics. The patient, surrounded by innovative AI tools, stands at the center of a complex system that captures various aspects of their condition and treatment data. The integration process predicts prognosis, leading to targeted care and potential re-optimization in cases of bad prognosis, thus marking a significant step toward more personalized, adaptable patient care.
© dkfz.de
Multi-Modal AI for RT Outcome Prediction
Personalized RT requires a profound understanding of the human body, a complex biological system with multifaceted characteristics that pose unique challenges in its modeling and analysis. To accurately depict this complexity, our approach utilizes a seamless integration of radiomics, dosiomics, and biomolecular omics through the use of state-of-the-art AI encoding and decoding architectures. Through this multi-omics convergence, we capture diverse biomarkers and physiological characteristics longitudinally over time and at specific instances. This data-driven method allows for the assessment of treatment response, enabling prognosis prediction, where subsequent analysis can aid in understanding what factors contributes to treatment outcomes and guides future RT strategies. Our framework represents a measured advancement in personalized RT, facilitating more precise and adaptable patient care and informing future treatment paths.
References
Salome, P., Sforazzini, F., Brugnara, G., Kudak, A., Dostal, M., Herold-Mende, C., Heiland, S., Debus, J., Abdollahi, A. and Knoll, M., 2023. MR-Class: A Python Tool for Brain MR Image Classification Utilizing One-vs-All DCNNs to Deal with the Open-Set Recognition Problem. Cancers, 15(6), p.1820.
Salome, P., Sforazzini, F., Kudak, A., König, L., Kickingereder, P., Bougatf, N., Wick, W., Jäkel, O., Debus, J., Knoll, M. and Abdollahi, A., 2021. Improved risk stratification via integration of radiomics and dosiomics features in patients with recurrent high-grade glioma undergoing carbon ion radiotherapy (CIRT).
Sforazzini, F., Salome, P., Kudak, A., Ulrich, M., Bougatf, N., Debus, J., Knoll, M. and Abdollahi, A., 2020. pyCuRT: An Automated Data Curation Workflow for Radiotherapy Big Data Analysis using Pythons' NyPipe. International Journal of Radiation Oncology, Biology, Physics, 108(3), p.e772.
Salome, P., Walz, D., Sforazzini, F., Kudak, A., Dostal, M., Regnery, S., Schlamp, K., Thomas, M., Herth, F., Jäkel, O. and Heußel, C.P., 2022. Multi-Omics Classifier of Tumor Recurrence vs. Radiation-Induced Lung Fibrosis in NSCLC Patients Treated with SBRT. International Journal of Radiation Oncology, Biology, Physics, 114(3), pp.e388-e389.
Salome, P., Sforazzini, F., Brugnara, G., Kudak, A., Dostal, M., Herold-Mende, C., Heiland, S., Debus, J., Abdollahi, A. and Knoll, M., 2023. MR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma. Cancers, 15(3), p.965.
Spadea, M.F., Pileggi, G., Zaffino, P., Salome, P., Catana, C., Izquierdo-Garcia, D., Amato, F. and Seco, J., 2019. Deep convolution neural network (DCNN) multiplane approach to synthetic CT generation from MR images—application in brain proton therapy. International Journal of Radiation Oncology* Biology* Physics, 105(3), pp.495-503.
Sforazzini, F., Salome, P., Moustafa, M., Zhou, C., Schwager, C., Rein, K., Bougatf, N., Kudak, A., Woodruff, H., Dubois, L. and Lambin, P., 2022. Deep Learning–based Automatic Lung Segmentation on Multiresolution CT Scans from Healthy and Fibrotic Lungs in Mice. Radiology: Artificial Intelligence, 4(2), p.e210095.
Contact and Funding
Contact: Dr. Patrick Salome
Funded by: Wilhelm Sander-Stiftung, funding number 3010001119