E5414. CT Perfusion Research: A Deep Dive via Topic Modeling
Authors
Burak Ozkara;
MD Anderson Cancer Center
Mert Karabacak;
Mount Sinai Health System
Konstantinos Margetis;
Mount Sinai Health System
Samir Dagher;
MD Anderson Cancer Center
Vivek Yedavali;
Johns Hopkins Hospital
Max Wintermark;
MD Anderson Cancer Center
Sotirios Bisdas;
University College London
Objective:
There is a demand for improved research synthesis methods that maximize efficiency. The topic modeling approach emerges as a novel solution to addressing this need. CT perfusion (CTP) has solidified its position as the preferred advanced imaging modality in many stroke clinical trials. Incorporating CTP into the standard operational procedures of stroke centers worldwide is becoming more common, demonstrating its effectiveness and dependability. CTP's growing popularity is also likely due to its high diagnostic accuracy in detecting myocardial ischemia. Given the rapid advancement and widespread adoption of CTP, we recognized the importance of launching a topic modeling project. Our study's goal is to gain an understanding of the dynamic research landscape surrounding CTP.
Materials and Methods:
To identify relevant articles, we searched the Scopus database from January 1, 2000 to August 16, 2023, using the keywords "computed tomography perfusion," "CT perfusion," "perfusion CT," and "perfusion computed tomography" in the article titles and keywords. Only "Article" and "Review" articles were considered. We used BERTopic, a topic modeling technique that aids in topic interpretation by preserving keywords in topic descriptions. Using the BERTopic model, we generated a set of topics and their corresponding representative documents. We also investigated the trends for the current decade.
Results:
The initial dataset had 3562 articles. After narrowing to "Article" and "Review" types, 995 were excluded and 53 more were excluded for lacking abstracts. Of the 2514 articles analyzed, 2356 were categorized into 12 groups. The remaining 158 were outliers, not fitting into any category. The topics that were crafted encompass tumor vascularity, stroke assessment, myocardial perfusion, intracerebral hemorrhage, imaging optimization, reperfusion therapy, postprocessing, carotid artery disease, seizures, hemorrhagic transformation, artificial intelligence, and Moyamoya disease. This chart captures the annual changes in topic prominence. The model offered insights into the current decade's patterns, spotlighting postprocessing and artificial intelligence as the most prominent subjects.
Conclusion:
Tumor vascularity was the most dominant topic. This area examines the importance of CTP in evaluating tumor vitality and vascularization, aiding in diagnosis, prognosis, and therapy response monitoring. Stroke assessment stood second, emphasizing CTP's role in acute stroke diagnosis significantly beyond the 6-hour mark. The increasing emphasis since 2018 is likely due to trials showing the efficacy of endovascular therapy in patients selected through CTP. Myocardial perfusion was the third prevalent topic, highlighting myocardial CTP imaging's accuracy and application in scenarios requiring clarity on ischemia presence. Lastly, the 2020s saw a focus shift towards postprocessing and artificial intelligence in CTP research, which are indicative of the challenges in standardizing postprocessing and the growing influence of artificial intelligence in radiology.