PhD Scientific Days 2022

Clinical Medicine VI. (Poster discussion will take place on the terrace of the room during the Coffee Break)

Pericardiac adipose tissue radiomics reveal an adverse fat pattern among heart failure patients in the UK Biobank imaging study

Előadó neve

Dr. Liliana, Szabo

Előadó munkahelye

Semmelweis University Heart and Vascular Center

Előadó telefonszáma

+36304843367

Előadó e-mail címe

sz.liliana.e@gmail.com

Az előadás címe

Pericardiac adipose tissue radiomics reveal an adverse fat pattern among heart failure patients in the UK Biobank imaging study

Szerző(k) neve és munkahelye

Liliana Szabo 1,2,3, Ahmed Salih 1,2, Esmeralda Ruiz Pujadas 4, Andrew Bard 1,2, Celeste McCracken 5, Hajnalka Vago 3, Bela Merkely 3, Stefan Neubauer5, Charalambos Antoniades 6, Karim Lekadir4, Steffen E. Petersen 1,2, Zahra Raisi-Estabragh 1,2

1. William Harvey Research Institute, NIHR Barts Biomedical Research Centre, Queen Mary University of London, Charterhouse Square, London, EC1M 6BQ, UK
2. Barts Heart Centre, St Bartholomew’s Hospital, Barts Health NHS Trust, West Smithfield, London, EC1A 7BE, UK
3. Semmelweis University, Heart and Vascular Center, Budapest, Hungary
4. Departament de Matematiques i Informatica, Universitat de Barcelona, Artificial Intelligence in Medicine Lab (BCN-AIM), Barcelona, Spain
5. Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, National Institute for Health Research Oxford Biomedical Research Centre, Oxford University Hospitals NHS Foundation Trust, Oxford, OX3 9DU, UK
6. Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, John Radcliffe Hospital, Headley Way, Oxford OX39DU, Oxford UK

Bemutatás módja

Poszter

Szekció

Clinical Medicine VI. (Poster discussion will take place on the terrace of the room during the Coffee Break)

Language of the presentation

English

Preferred session

Clinical Medicine

Összefoglaló szövege

Background: Obesity is a rising public health crisis. It is increasingly recognized that different fat compartments have different risk relationships. Fat accumulation, specifically disproportionate deposition of metabolically active visceral and pericardial adipose tissue (PAT), is linked to an increased hazard of obesity-related cardiovascular outcomes. Furthermore, current data suggest a distinct mechanistic role for PAT in driving adverse cardiovascular outcomes. Existing work is limited to describing PAT quantity. However, the character of PAT also has likely mechanistic importance.

Aims: In this study, we use radiomics measures of PAT geometry and tissue character derived from pixel-level magnetic resonance imaging to define imaging signatures of PAT which discriminate heart failure (HF).

Methods and results: We studied 43,226 UK Biobank participants with cardiovascular magnetic resonance (CMR), of these 297 had prevalent HF at the time of imaging. We obtained PAT contours using an automated pipeline. The PyRadiomics platform was adopted to extract shape, first-order and texture features (n=105). To create a balanced cohort, we randomly matched equal number of individuals without HF as comparator group (n=297). The data were split into training and test sets. (20%). We applied 6 classification models with 10-folds cross-validation and hyperparameter tuning. All sex and age-adjusted classification models achieved good discrimination between HF patients and controls. The best discriminative power was reached using the voting classifier method (Accuracy: 0.79, F1 score: 0.79). We applied SHAP to describe the feature importance of PAT radiomics parameters. We found that texture (e.g. High Gray level emphasis) and shape features were the most important metrics distinguishing prevalent HF.

Conclusion: We demonstrated that HF patients have distinct radiomics PAT phenotypes with important information from both shape and texture features. Our findings provide new insights into mechanisms of PAT-HF associations and propose novel imaging biomarkers for their investigation.

Funding: LS was supported by the EFOP-3.6.3-VEKOP-16-2017-00009.

University and Doctoral School

Semmelweis University, Doctoral School of Theoretical and Translational Medicine

Other university and doctoral school, not listed above

William Harvey Research Institute, NIHR Barts Biomedical Research Centre, Queen Mary University of London, Charterhouse Square, London, EC1M 6BQ, UK

Supervisor

Hajnalka Vago and Zahra Raisi-Estabragh

Publication of my abstract

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

Kind

Szabad

Status

elfogadva

Accepted presentation method

poszter

Előadás fájl jóváhagyás

nem rendelkezett róla

Előadó

4153

Start

11:10

End

11:15

Authors (legacy)

Liliana Szabo 1,2,3, Ahmed Salih 1,2, Esmeralda Ruiz Pujadas 4, Andrew Bard 1,2, Celeste McCracken 5, Hajnalka Vago 3, Bela Merkely 3, Stefan Neubauer5, Charalambos Antoniades 6, Karim Lekadir4, Steffen E. Petersen 1,2, Zahra Raisi-Estabragh 1,2

1. William Harvey Research Institute, NIHR Barts Biomedical Research Centre, Queen Mary University of London, Charterhouse Square, London, EC1M 6BQ, UK
2. Barts Heart Centre, St Bartholomew’s Hospital, Barts Health NHS Trust, West Smithfield, London, EC1A 7BE, UK
3. Semmelweis University, Heart and Vascular Center, Budapest, Hungary
4. Departament de Matematiques i Informatica, Universitat de Barcelona, Artificial Intelligence in Medicine Lab (BCN-AIM), Barcelona, Spain
5. Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, National Institute for Health Research Oxford Biomedical Research Centre, Oxford University Hospitals NHS Foundation Trust, Oxford, OX3 9DU, UK
6. Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, John Radcliffe Hospital, Headley Way, Oxford OX39DU, Oxford UK