CORTEXA
← Browse
zenodoConference paper2026-05-20

SoC-Based Implementation of CNN Model for End-Diastolic Volume Classification from Echocardiogram via hls4ml

Nemanja Marković, Tijana Geroski, Emil Jovanov, Nenad Filipović

Echocardiographic assessment of End-Diastolic Volume (EDV) is central to identifying dilated cardiomyopathy, a major driver of heart failure. This paper presents a low-latency, edge-computing solution that deploys an 8-bit quantized, 50%-pruned CNN directly onto a Xilinx Artix-7 FPGA via the hls4ml interface, enabling real-time, on-device classification of EDV/ventricular enlargement from echocardiogram frames without relying on cloud infrastructure. The approach is designed for integration into portable ultrasound devices to support point-of-care heart failure screening in emergency departments and resource-limited clinics. This work was presented at the 5th Serbian International Conference on Applied Artificial Intelligence (SICAAI 2026), Kragujevac, Serbia, and was carried out within the STRATIFYHF project.

View free PDFSource page

Related papers

zenodoConference paper2026-07-28

Self-Supervised Relevance Modelling in Autonomous Driving via Counterfactual Analysis

Luca Lusvarghi, Javier Gozalvez, Pablo Urbano Hidalgo

Autonomous driving relies on computationally intensive perception pipelines to continuously detect and track objects in the surrounding environment. While some objects are key to plan safe and effective maneuvers, others may not be relevant and have no impact on the autonomous ve…

View free PDFSource page
zenodoConference paper2026-05-20

Echocardiography View Classification as an Important Step to Heart Failure Diagnosis: A Case Study of the EchoJEPA Foundation Model

Ognjen Pavić, Lazar Dašić, Tijana Geroski, Anđela Blagojević, Andrej Preveden, Aleksandra Milovančev, et al.

Accurate classification of echocardiographic views (e.g. apical 2- and 4-chamber, parasternal long-axis) is an important prerequisite for reliable ejection-fraction assessment and heart failure diagnosis, but manual classification is time-consuming. This study evaluates the EchoJ…

View free PDFSource page
zenodoConference paper2026-05-20

Comparative Analysis of U-Net-Based Architectures for Coronary Artery Segmentation Using X-Ray Angiography Images

Anđela Stojadinović, Tijana Geroski, Dajana Jovanović, Nenad Filipović

Accurate delineation of coronary arteries from X-ray angiography supports early identification of narrowing or blockages, but is complicated by low contrast and thin, branching vessel structures. This study compares three U-Net-based segmentation architectures — U-Net, U-Ne…

View free PDFSource page
zenodoConference paper2025-04-08

Risk Stratification and Early Diagnosis of Heart Failure

Borut Flis, Petar Vračar, Matej Pičulin, Djordje Jakovljević, Nenad Filipović, Zoran Bosnić

Heart failure (HF) affects over 64.3 million people worldwide. As a part of the StratifyHF project, we developed a decision support system (DSS) to enhance HF prediction and diagnosis through machine learning (ML) approaches. The DSS comprises two modules: Early diagnosis and Ris…

View free PDFSource page
zenodoConference paper2025-05-29

Assessment of Morbidity in Patients with Heart Failure Using Traditional Machine Learning Techniques

Đorđe Ilić, Ognjen Pavić, Lazar Dašić, Anđela Blagojević, Nenad Filipović

Morbidity and disease severity in heart failure are commonly assessed using New York Heart Association (NYHA) classes. This study uses non-invasive data — physical examination, symptoms and disease history — to classify patients into four morbidity classes approximati…

View free PDFSource page
zenodoConference paper2026-02-15

A Modular Deep Learning Pipeline for Echocardiographic Video Classification of Heart Failure

Tijana Geroski, Mahyar Bolhassani, Narayana S. Singam, Ayman Battisha, Nenad Filipović, Dinesh Kalra, et al.

Distinguishing heart failure with reduced ejection fraction (HFrEF) from heart failure with preserved ejection fraction (HFpEF) is clinically important but challenging. This paper presents an end-to-end deep learning pipeline for automated three-class classification (healthy, HFr…

View free PDFSource page