All publications

Conference paper / 2023

CenterlinePointNet++: A New Point Cloud Based Architecture for Coronary Artery Pressure Drop and vFFR Estimation

Patryk Rygiel, Paweł Płuszka, Maciej Zięba, Tomasz Konopczyński

Medical Image Computing and Computer Assisted Intervention (MICCAI) · Published

CenterlinePointNet++ uses a surface point cloud and centerline with encoder and decoder blocks to predict coronary hemodynamics.
CenterlinePointNet++ combines the vessel surface and centerline graph. Architecture diagram reproduced from the authors’ ML in PL 2023 presentation. Architecture diagram · Rygiel et al. (2023), slide 15 · © 2023 Hemolens Diagnostics Sp. z o.o. · authors’ conference slides

Learning along the vessel tree

Nearby points in Euclidean space do not necessarily belong to the same artery branch. CenterlinePointNet++ incorporates vessel-centerline information into point-cloud processing to better reflect the structure of the coronary tree.

The model predicts pressure drop and virtual fractional flow reserve from vessel geometry. Its evaluation explores neural estimation as an alternative to running a full flow simulation for each new case.

The accompanying MSc thesis provides a longer account of this research.

Presented as a poster at MICCAI 2023 in Vancouver. See the conference programme.

Cite this work

@inproceedings{centerlinepointnet2023,
  title = {CenterlinePointNet++: A New Point Cloud Based Architecture for Coronary Artery Pressure Drop and vFFR Estimation},
  author = { Patryk Rygiel and Paweł Płuszka and Maciej Zięba and Tomasz Konopczyński },
  year = {2023},
  booktitle = {Medical Image Computing and Computer Assisted Intervention (MICCAI)},
  pages = {781--790},
  doi = {10.1007/978-3-031-43990-2_73},
  url = { https://doi.org/10.1007/978-3-031-43990-2_73 }
}