Electronic Thesis/Dissertation
 

Radiomics-Driven Predictive Lung Cancer Models

Open Access Deposited

Despite advancements in medical imaging technologies, accurately classifying subtypes of non-small cell lung cancer (NSCLC) remains a significant diagnostic challenge. This difficulty often leads to treatment delays and leads to less effective care. This praxis takes a different approach. It investigates a radiomics-driven approach to address this gap. This approach leverages quantitative imaging features and machine learning algorithms to improve subtype classification accuracy.The study focused on the three main subtypes of NSCLC. These subtypes are adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. A total of 855 handcrafted radiomics features were extracted from segmented CT scans. These features capture detailed information about the tumor's shape, texture, and intensity, offering insights that are not visible to the human eye. Several supervised machine learning models were trained using these features. These models include support vector machines, random forests, XGBoost, and a stacking ensemble. Publicly available datasets were used for training. Performance was evaluated through multiple metrics such as precision, recall, F1-score, confusion matrices, and ROC curves. Among all the models tested, the stacking ensemble delivered the strongest performance. It achieved an overall accuracy of 81% and maintained balanced classification across all three subtypes. These results show that radiomics features contain meaningful information capable of supporting reliable multiclass classification, even without the use of deep learning. This work offers a reproducible framework with potential clinical relevance. Its emphasis on model interpretability makes it easier for healthcare providers to understand and apply the results. By moving beyond simple malignant or benign distinctions, this research demonstrates how radiomics and machine learning can contribute to more personalized and precise lung cancer diagnosis. These findings open the door for further research and potential improvements in patient care.

Author Language Keyword Date created Type of Work License
  • All rights reserved
Rights statement GW Unit Degree Advisor Committee Member(s) Persistent URL

Notice to Authors

If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.

Thumbnail Title Date Uploaded Visibility Actions
Preview of Essien_gwu_0075A_17397.pdf Essien_gwu_0075A_17397.pdf 2025-12-11 Open Access