Automatic segmentation of Doppler echocardiographic flow traces with deep learning

Title:        Automatic segmentation of Doppler echocardiographic flow traces with deep learning.

Team:        MYRIAD = « Modeling & analysis for medical imaging and Diagnosis »

Supervision:    Nicolas DUCHATEAU (Full Professor) - Olivier BERNARD (Full Professor)

Context:     Cardiac diseases progressively deteriorate the cardiac function, including myocardial wall shape and deformation, and flow across the different valves (for the left ventricle: mitral and aortic). Doppler echocardiographic allow visualizing flow patterns and relating them with the known disease characteristics. However, their quantification is challenging and remains limited to simple measurements (peak and slope values, timings) with very limited automation [JEV-22, AGU-24], which hampers their inclusion in large population analyses.

We have developed strong expertise on the automatic analysis of 2D echocardiography, mostly regarding segmentation and tracking of the left ventricle [JUD-26a, JUD-26b]. Through this internship, we would like to complement the analysis of such data with flow traces automatically extracted from the corresponding Doppler acquisitions.

Figure: adapted from [LAN-16]. Top: Representative mitral flow patterns depending on disease. Bottom: Example of a mitral flow Doppler acquisition.

Objectives:     This internship will specifically target:
- Designing efficient and robust automatic segmentation of Doppler traces, mostly for mitral and aortic flow,
- Improving its physiological and temporal consistency,
- Classifying diseases based on the extracted Doppler traces.
- Evaluating the segmentation and classification performance on existing 2D echocardiographic studies.

Practical information:
   • Location: DOUA campus, CREATIS lab, Villeurbanne.
   • Duration: 6 months, starting February-March 2027.
   • Exploratory subject in close collaboration with other researchers focusing on segmentation and tracking.

Profile:
   • MSc student with an applied mathematics and/or computer science background. 
   • Good programming and image processing skills in Python.
   • Good English
   • Motivated to work on medical applications.

Contact:    Send your CV, recommendation letter and/or coordinates of referent professors, and academic record to:
    nicolas.duchateau@creatis.insa-lyon.fr

Bibliography:

[AGU-24] Aguado et al. Front Digit Health 2024;6:1455767.
[JEV-22] Jevsikov et al. MIDL 2022.
[JUD-26a] Judge et al. IEEE TMI 2026;45:4161-73.
[JUD-26b] Judge et al. MICCAI-STACOM 2026.
[LAN-16] Lancellotti & Cosyns. The EACVI echo handbook 2016.