AI-Driven Volumetric 17-Segment Myocardial Blood Flow Characterisation Using [82Rb] PET/CT
| Type of publication: | Misc |
| Citation: | |
| Year: | 2026 |
| Month: | October |
| Howpublished: | ePoster at the Annual Congress of the European Association of Nuclear Medicine (EANM) |
| Abstract: | Aim/Introduction: Dynamic 82Rb PET/CT enables quantitative assessment of myocardial blood flow (MBF), which has important diagnostic and prognostic value. However, robust segment-level MBF analysis requires reliable left ventricular (LV) segmentation and regional assignment. We developed an AI pipeline for 17-segment MBF characterisation from 82Rb PET. Materials and Methods: 50 patients (17 women, 33 men) with high ischemia and necrosis (median 4.4% and 14.7%) who underwent rest and stress 82Rb PET/CT were manually delineated to obtain LV ground truth (GT). The cohort was processed with FlowQuant to obtain standard-of-care MBF, arterial input function (AIF) and 17-segment regional mapping. The mapping was adapted to cover the whole heart and applied to the GT to obtain the segmental GT, which is needed due to partial exclusion of basal voxels due to approximate fitting in standard-of-care. We trained nnU-Net models following a stratified 5-fold cross-validation scheme using two early and two late frames to automatically segment the 17 segments and extract their centrelines (LVCL). Volumetric MBF was obtained using AI-enabled 3D mapping [1], further sampled on LVCL using the segmental prediction and averaged per region MBFReg(i). Conclusion: Our fully automatic segmentation and CL extraction pipeline on 82Rb PET, using semi-automatic AIF, allows volumetric segmental MBF characterisation aligned with standard-of-care. |
| Keywords: | Myocardial Blood Flow, Rubidium PET, Transmural Perfusion |
| Authors | |
| Added by: | [] |
| Total mark: | 0 |
|
Attachments
|
|
|
|
|
Notes
|
|
|
|
|
|
Topics
|
|
|
|
