Brancato, Valentina (2024) Computational Approaches to Support Clinical Decision on Different Scales of Biomedical Images in Oncology. [Tesi di dottorato]

Anteprima

Testo Brancato_Valentina_36.pdf Download (108MB) | Anteprima

Abstract

In the era of precision medicine, the integration of heterogeneous data across multiple scales is crucial for advancing cancer diagnosis and prognosis. Despite significant advancements in diagnostic techniques and computational analysis led to a big data production, these data are often characterized by a high heterogeneity and complexity, belonging to different domains, and their integration remains an open challenge. This PhD thesis explores a possible solution through Computational Decision Support Systems employing AI and data analytics for aggregating, structuring, and comprehending biomedical data, with a focus on medical imaging, that have historically played a crucial role in cancer screening, diagnosis, staging, and therapeutic response monitoring. The thesis first emphasizes the urgency of organizing complex biomedical data belonging to different diagnostic domains, framing digital biobanks as a multifactorial solution capable of containing curated and standardized imaging data, along with clinical, molecular, and pathologic data. Based on this foundation, the thesis delves into the development of computational and statistical tools for analyzing biomedical images across different imaging scales. The correlation between radiomic and pathomic features was explored, particularly in the context of Glioblastoma Multiforme. Preliminary findings revealed intriguing cross-scale relationships, offering a nuanced understanding of tumor heterogeneity and impacting diagnostic, prognostic, and therapeutic considerations. The concept of a "virtual biopsy" emerges, representing a transformative shift in diagnostic methods by relying on advanced analyses of radiological images. Furthermore, the thesis demonstrates the practical applications of pathomics in cancer diagnosis, exemplified through studies in breast and prostate cancers. The pathomic approach proved valuable in quantifying tumor-infiltrating lymphocytes in breast cancer and improving Gleason grading in prostate cancer. These applications showcase the transformative potential of pathomics in enhancing diagnostic precision and treatment strategies, positioning it as a key player in the future of oncology. In summary, this PhD thesis makes significant contributions to the field by establishing a foundation for advancing precision medicine through computational approaches in oncological decision-making. The proposed model for digital biobanks, the exploration of cross-scale relationships in glioblastoma, and the practical applications of pathomics collectively contribute to the development of comprehensive tools that have the potential to revolutionize cancer diagnosis and decision support. This work serves as a compelling call to action for the scientific community to embrace these innovative approaches and drive positive changes in clinical practice, ultimately improving patient outcomes in oncology.

Downloads

Actions (login required)

Modifica documento