PhD Scientific Days 2025

Cardiovascular Medicine and Research I.

Infarctsize-AI: An efficient preclinical infarct size image analysis tool

Előadó neve

Dr. Kovácsházi Csenger

Neptun code

E31RMH

Előadó munkahelye

Semmelweis University

Előadó telefonszáma

+36 1 459 1500/56264

Előadó e-mail címe

kovacshazi.csenger@semmelweis.hu

Az előadás címe

Infarctsize-AI: An efficient preclinical infarct size image analysis tool

Szerző(k) neve és munkahelye

Csenger Kovácsházi¹, Dóra Kapui², Bennet Y Weber², Tamás G Gergely², Gábor B Brenner², Bence Ágg², Csanád Tabajdi³, Adrienn Rácz³, András Horváth³, Sauri Hernandez-Resendiz⁴, Derek J Hausenloy⁴, Reinis Vilskersts⁵, Marta Oknińska⁶, Michał Waszkiewicz⁶, Michal Mączewski⁶, Arnold Molnár⁷, Tamara Szabados⁷, Péter Bencsik⁷, Thomas Krieg⁸, Javier Inserte⁹, Rainer Schulz¹⁰, Coert J Zuurbier¹¹, Ioanna Andreadou¹², Bruno K Podesser¹³, Péter Ferdinandy¹, Zoltán Giricz¹

1: Department of Pharmacology and Pharmacotherapy, Semmelweis University, Budapest, Hungary.
2: Center for Pharmacology and Drug Research & Development, Semmelweis University, Budapest, Hungary.
3: Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest, Hungary
4: Cardiovascular and Metabolic Disorders Programme, Duke-NUS Medical School, Singapore, Singapore
5: Latvian Institute of Organic Synthesis, Riga, Latvia
6: Department of Clinical Physiology, Centre of Postgraduate Medical Education, Warsaw, Poland
7: Cardiovascular Research Group, Department of Pharmacology and Pharmacotherapy, Albert Szent-Györgyi Medical School, University of Szeged, Szeged, Hungary.
8: Department of Medicine, University of Cambridge, Cambridge, United Kingdom
9: Department of Cardiology, Vall d'Hebron University Hospital and Research Institute, Universitat Autònoma, Barcelona, Spain.
10: Institute of Physiology, Justus Liebig University Giessen, Giessen, Germany
11: Laboratory of Experimental Intensive Care and Anesthesiology (L.E.I.C.A.), Department of Anesthesiology, Amsterdam Cardiovascular Sciences, Amsterdam UMC, University of Amsterdam, The Netherlands
12: National and Kapodistrian University of Athens, Faculty of Pharmacy, Department of Pharmaceutical Chemistry, Laboratory of Pharmacology, Panepistimiopolis, Zografou, Athens, Greece
13: Ludwig Boltzmann Institute for Cardiovascular Research, Center for Biomedical Research, Medical University of Vienna, Vienna, Austria

Bemutatás módja

Szóbeli

Szekció

Cardiovascular Medicine and Research I.

Language of the presentation

English

Preferred session

Cardiovascular Medicine and Research

Összefoglaló szövege

Myocardial infarct size (IS) is a leading predictor of adverse cardiac outcomes following acute myocardial infarction (AMI). However, IS quantification in preclinical models is time-consuming and prone to inter-observer variance. To address these challenges, we developed Infarctsize-AI, an artificial intelligence (AI)-based application to enhance speed, and reproducibility while reducing analysis bias and workload of preclinical IS analysis.
Images of rat and mouse heart slices stained with Evans Blue, to define area at risk (AAR), and 2,3,5-triphenyltetrazolium chloride (TTC), to identify IS, from pre-existing research were used. Data of three study centers were utilized to train deep learning image segmentation models for IS analysis. Data from one of these centers (dependent data) and from independent sources (independent data) were used to test AI pre-annotation.
We observed strong agreement in IS/AAR between AI pre-annotation and AI-assisted annotation, consisting of AI pre-annotation and its supervision on dependent data (mean difference = 2.3%, 95% limits of agreement [-7.92%,12.5%]), however, in 7.32% of the cases, difference over the pre-defined 10% cutoff value were identified. On independent data, AI successfully outlined the slices and AAR but failed to delineate infarct regions. Besides, a Confidence Score was developed to flag AI pre-annotations with low reliability. The score effectively highlighted regions with overlap between AI pre-annotation and manual annotation below 75% (area under receiver operator characteristics curve for AAR: 0.943, infarction: 0.926). Furthermore, analysis time was compared between AI-assisted and manual annotations. AI-assisted annotation reduced analysis time by 70.8% compared to manual annotation.
Infarctsize-AI (available at: infarctsize.com), offers an AI-driven solution for preclinical IS quantification, streamlining image analysis, reducing time, and standardizing annotation documentation to improve efficiency and reproducibility.

University

Semmelweis University

Supervisor

Dr. Zoltán Giricz

Publication of my abstract

I do not give consent to the publication of my abstract on the website of the congress.

phd.section.field

after finishing doctoral studies with absolutorium (PhD)

Kind

Szabad

Status

elfogadva

Accepted presentation method

szóbeli

Előadás fájl jóváhagyás

nem hagyta jóvá

Előadó

5976

Start

12:30

End

12:45