Comprehensive Tissue characterization with deep learning models utilizing NLST data for chest CT analysis
By leveraging the NLST's extensive chest CT dataset, which includes serial scans from a large and diverse population, we aim to refine and validate accurate models for tissue characterization. These models will enable us to monitor tissue changes over time with greater precision, providing valuable insights into disease progression and treatment effects. The diversity of the NLST data enhances the generalizability of our tools, ensuring that the AI-models are applicable to a wide range of populations.
By the end of this project, we expect to deliver a state-of-the-art AI tool that significantly improves personalized tissue characterization from standard chest CT, contributing to better patient outcomes and advancing the field of medical imaging.
Our project encompasses three primary aims, which will result in tools for comprehensive tissue characterization from chest CT scans.
Aim 1: Refinement of Tissue Characterization Models: We intend to refine the existing deep-learning models that segment different body structures and differentiate various types of tissues. By leveraging the extensive NLST dataset, we will validate these models across different sex and race groups to ensure their accuracy and generalizability.
Aim 2: Evaluation of Tissue Characterization from Chest CT Scans: Building upon our refined models, we will analyze tissue characteristics in various organs. Recognizing that variations in scan acquisition protocols can affect measurements, we plan to develop custom approaches tailored to specific protocols.
Aim 3: Detection of Tissue Parameter Changes Over Multiple Scans: We will assess changes in tissue parameters over time by analyzing serial chest CT scans from the NLST dataset. Our AI-based models will detect and quantify these changes, which will then be evaluated by radiologists to validate our approaches.
Throughout the study, we will conduct sex- and race-specific analyses to address potential disparities and ensure that our models are applicable to diverse populations. The successful completion of these aims will allow deployment of methods for comprehensive tissue characterization using chest CT data.
Members of the R&D team of APQ Health Inc.,