# Characterizing longitudinal dynamics of indeterminate pulmonary nodules using semantic, radiomic, and deep features > **NIH NIH U01** · UNIVERSITY OF CALIFORNIA LOS ANGELES · 2024 · $138,972 ## Abstract This application is being submitted in response to the Notice of Special Interest (NOSI) defined as “NOT-CA-24-058”. Landmark studies such as the National Lung Screening Trial (NLST) demonstrated that low-dose computed tomography (LDCT) screening reduces lung cancer mortality, but the growing use of LDCT has also led to the annual discovery of over a million indeterminate pulmonary nodules (IPNs), which in the context of this project are solid noncalcified nodules between 6-30 mm in diameter. While IPNs are largely benign, their discovery induces anxiety among patients, creates a burden on the healthcare system, and, for some patients, represents early-stage cancer. Newer imaging-based biomarkers and the use of the patient’s evolving medical history can improve prediction and support earlier diagnosis in the setting of the IPNs. Key challenges to overcome include handling variations in how CT studies are acquired, characterizing and modeling changes in nodular and perinodular features observed over longitudinal imaging scans, and predicting whether an IPN is malignant. This supplement utilizes longitudinal clinical and imaging data collected as part of the NLST but addresses several critical limitations: (1) nodules were not uniquely identified, (2) nodules were not tracked over multiple scans, (3) the precise location of nodules and their boundaries were not delineated, and (4) lung cancer diagnoses were not linked to individual nodules. This project complements the parent U01 (EFIRM Liquid Biopsy Research Laboratory: Early Lung Cancer Assessment, U01 CA233370, Contact PI: Wong), which seeks to develop and validate a blood-based technology to distinguish benign and malignant IPNs. Our overarching hypothesis is that the combination of longitudinal semantic (radiologist-interpreted), radiomic (e.g., handcrafted shape, intensity, and texture), and deep (e.g., neural-network learned) features can predict nodule malignancy among IPNs with higher positive predictive value than nodule volume alone. We will investigate whether the change in these features (delta image features) captures the natural history of tumor and peritumoral appearance, hypothesizing that these subtle changes reflect shifts in the tissue microenvironment and can be used to di]erentiate between benign and malignant nodules. We accomplish this with two aims: (1) characterize longitudinal changes in nodules using semantic, radiomic, and deep features and (2) assess the predictive value of longitudinal features in classifying IPNs. We have established a team of investigators with expertise in liquid biopsy, machine learning/informatics, and thoracic radiology. Our e]orts will collectively result in a comprehensive annotated dataset of NLST IPN cases as they evolve and the identification of delta image features that can be surrogate endpoints for clinical trials and physician support services. ## Key facts - **NIH application ID:** 11135001 - **Project number:** 3U01CA233370-07S1 - **Recipient organization:** UNIVERSITY OF CALIFORNIA LOS ANGELES - **Principal Investigator:** William Hsu - **Activity code:** U01 (R01, R21, SBIR, etc.) - **Funding institute:** NIH - **Fiscal year:** 2024 - **Award amount:** $138,972 - **Award type:** 3 - **Project period:** 2018-09-13 → 2025-08-31 ## Primary source NIH RePORTER: https://reporter.nih.gov/project-details/11135001 ## Citation > US National Institutes of Health, RePORTER application 11135001, Characterizing longitudinal dynamics of indeterminate pulmonary nodules using semantic, radiomic, and deep features (3U01CA233370-07S1). Retrieved via AI Analytics 2026-06-03 from https://api.ai-analytics.org/grant/nih/11135001. Licensed CC0. --- *[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*