AAI_2025_Capstone_Chronicles_Combined

14

perspective as both a classmate and subject matter expert has been invaluable. I would also like to acknowledge the NSCLC-Radiomics (Lung1) dataset contributors and The Cancer Imaging Archive (TCIA) for creating and maintaining a clean, well-curated, and rigorously annotated public dataset. Their work in assembling high-quality imaging data and clinically reviewed tumor contours makes reproducible research in lung cancer imaging possible and directly enabled the research presented in this project. Disclosures I completed this project as an independent graduate student researcher. The work presented here was not conducted on behalf of, funded by, or reviewed by any employer or past employer. Although I previously served as Senior Director of Software Engineering at BillionToOne (Menlo Park, CA), a clinical diagnostics company, that role did not involve work related to lung cancer imaging, radiotherapy planning, or volumetric tumor assessment. No proprietary data, models, software, or techniques from BillionToOne - or any other prior employer - were used in this project. All analyses rely exclusively on publicly available datasets and open-source tools. I hold several prior leadership and advisory roles in healthcare-adjacent technology organizations, including Kaia Health and Durg.ai. None of these organizations develop tools for radiologic segmentation or volumetric tumor measurement, and none contributed resources, funding, or technical assets to this research. I am the named inventor on multiple U.S. patents related to information mining and predictive modeling. These patents are unrelated

to medical imaging or the methods used in this study. There are no financial conflicts of interest to declare. No external grants, industry partnerships, or commercial entities provided

support for this project. AI Use Disclosure

This project was conceived, designed, and executed by me, drawing on my own technical judgment, research experience, and subject-matter expertise in healthcare, medical imaging, and machine learning. AI-based tools - including OpenAI’s ChatGPT and the Cursor AI coding environment - were used as assistive resources to support my workflow, not to replace my own reasoning or authorship. ChatGPT was used to help identify relevant peer-reviewed literature, verify that claims in the manuscript were appropriately supported, refine clarity and academic tone in my writing, and provide guidance on organization and formatting. Cursor was used to assist in writing portions of the project’s code. In all cases, I made the technical decisions about software design, algorithmic approach, project structure, and coding style. Any code generated with AI assistance was exhaustively reviewed, tested, modified, and validated by me to ensure correctness, alignment with project requirements, and adherence to best practices. All conceptual ideas, methodological choices, data processing logic, experimental design, evaluation strategy, and interpretation of findings are fully my own. The AI-assisted components served only to enhance efficiency and clarity. This disclosure is provided in accordance with university guidelines requiring explicit acknowledgment of AI tools and

411