Artificial Intelligence Imaging Research Programs
The Artificial Intelligence Imaging Research Center is currently working on four well-defined, multi-disciplinary image-based AI research projects. All projects are clinical in nature and in part promote standardization.
Creation of a Modernized Automated Bone Age Interpretation Tool
Radiographic bone age assessment is used in numerous settings to guide medical and surgical treatment. We are creating a modernized automated bone age interpretation tool based on a library of more than 20,000 normal trauma hand radiographs acquired at Cincinnati Children's Hospital Medical Center. Such an algorithm will considerably improve bone age estimation accuracy as well as account for modern bone maturation, while decreasing the inter-radiologist variability.
Automated Radiographic Lines, Drains, and Airways (LDAs) Detection in Children
Lines, Drains, and Airways (LDAs) are an essential part of complex post-operative care for surgical patients who undergo operative procedures. It is critically important to evaluate the presence and locations of lines (e.g., vascular catheters), drains (e.g., chest tubes, mediastinal drains), and airways (e.g., endotracheal tubes, tracheostomy tubes) using radiographic images. The overarching objective of this multi-disciplinary project is to create an object identification algorithm that will identify the presence and locations of clinically important LDAs based on pediatric chest radiographs at Cincinnati Children’s.
Automated Measurement of Histologic Liver Fibrosis in Children and Young Adults
Histologic assessment of core biopsy specimens is the current reference standard for detecting and quantifying the presence and severity of liver fibrosis. Fibrosis staging may use one of several semi-quantitative systems. The objective of this project is to create an AI model to perform automated quantification (both quantitative and semi-quantitative) of histologic liver fibrosis in children with pediatric chronic liver diseases.
Automated CT/MRI Organ Segmentation
Automated segmentation of organs and tissues from cross-sectional CT and MRI images has a variety of clinical and research applications. The objectives of this project include: 1) Develop and deploy automatic organ segmentation models for liver, spleen, pancreas, kidneys, and skeletal muscle for common CT and MRI examinations; 2) Develop and deploy Total Cardiac Volume segmentation model for pediatric transplant patients imaged using CT and create a framework for the multi-institutional application; 3) Develop automatic segmentation models of lung and airway for pediatric NICU and obstructive sleep apnea populations scanned using MRI.
Associated Research Labs
Cincinnati Children’s has one of the largest repositories of clinical digital pediatric imaging data (e.g., radiology, cardiology, histology, endoscopy, etc.) and clinical data (e.g., laboratory, “-omics” data) in the world. Our research labs have been dedicated to exploring the frontiers of imaging-based research using cutting edge AI techniques.
