A CT imaging-based AI framework can boost opportunistic screening for splenomegaly, suggest findings presented October 21 at the Conference on Machine Intelligence in Medical Imaging (CMIMI).
In his talk, delivered at the Society for Imaging Informatics in Medicine (SIIM)-hosted meeting, MD/PhD candidate David Zhang of the University of Pennsylvania in Philadelphia discussed the success of his team’s framework, which he highlighted was validated with physician-determined large spleens in a clinical population with systemic multiorgan diseases.
“There were a lot of cases of splenomegaly that our model may have identified, but the patients themselves might not have previously categorized as having,” Zhang told AuntMinnie.com.
Splenomegaly is the enlargement of splenic size and weight, which can lead to disruption of the spleen’s role in immunological defense and hematopoiesis, and CT imaging is the go-to modality for diagnosing the disease.
Opportunistic CT screening gathers information about clinical conditions when people are imaged for other reasons and can improve the radiology workflow with detailed quantitative imaging features from radiomics, Zhang explained. He and his colleagues built and deployed an end-to-end opportunistic screening workflow called PheWAS, which consists of AI-based automated image analysis embedded in the radiology clinical workflow. They validated the system by measuring spleen volumes and showed associations with systemic multiorgan diseases using a phenome-wide association study.
“We have a lot of clinical imaging data and clinical data in general, and we’d love to figure out how we can integrate the two and potentially use the imaging data to help better enforce and define our clinical diagnoses,” Zhang said.
The team estimated spleen volumes for 13,636 people who underwent CT imaging in 2023 and 2024 using an online segmentation tool called TotalSegmentator.
It also determined splenomegaly assignments if a patient had an International Classification of Diseases (ICD) diagnosis for the condition prior to imaging scans. It also performed a phenome-wide association study adjusting for sex, age, height, and body mass index (BMI).
The investigators found that based spleen volume measurements from PheWAS were 484.3 mL for splenomegaly and 202.5 mL for other patients.
Zhang said that increased spleen volumes were strongly tied to over 50 clinical entities including digestive system diseases and multiorgan diseases. These included infections, endocrine disease, and kidney disease.
Zhang highlighted that based on these results, PheWAS can be an effective tool for opportunistic screening of suspected splenomegaly and other conditions.
“This could be applied to a variety of other conditions … to help radiologists and clinicians help diagnose conditions or at least put conditions on their radar … that they can take into account when making clinical decisions,” he told AuntMinnie.com.
Check out AuntMinnie.com's coverage of CMIMI 2024 here.
