1)The group has continued to work on developing an atlas of fibre orientations using high resolution (submillimetre) diffusion tensor images of explanted porcine hearts. Mia Mojica (PhD student at UOIT) presented progress on group-wise registration and diffusion tensor re-orientation at the Imaging Network Ontario Symposium (Toronto, March 2018). An extended work on building the DTI-based atlas was submitted in August 2018 and accepted at the next international SPIE medical imaging conference (San Diego 2019) “Constructing an average geometry and diffusion tensor magnetic resonance field from freshly explanted porcine hearts”, Mia Mojica, Mihaela Pop, Maxime Sermesant, Mehran Ebrahimi.
2)The group has also continued to refine their 3D MRI-based computational models for electrophysiology simulations. Several heart models and meshes have been built from in vivo T1-maping images of pre-clinical infarcted pig hearts and used for testing various mathematical reaction-diffusion models (Aliev-Panfilov, Eikonal, Mitchel-Schaeffer). Simulations were performed using synthetic (rule-based) fibres. Future work (next year) will focus on comparisons between simulated activation maps obtained from models integrating real DTI-based fibre directions vs synthetic fibres vs atlas fibres. Mengyuan Li (Sunnybrook student) working with supervisor Dr Mihaela Pop (Sunnybrook-PI) presented results at the 62ndBiophysical Society international meeting (San Francisco February 2018), Imaging Network Ontario (Toronto, March 2018), and she was also awarded a Honourable mention at Sunnybrook Research student competition with the poster entitled “Novel T1 mapping-based pre-clinical models for cardiac EP: a combined experimental and theoretical study”: Mengyuan Li, Maxime Sermesant, Fumin Guo, Peter Lin, Seb Ferguson, Jen Barry, Graham Wright, Mihaela Pop.
3)Inria has also progressed with the fast simulation of cardiac electrophysiology, based on CT imaging and the Eikonal model from clinical application, with now a publication in a clinical journal: « Fast Personalized Electrophysiological Models from CT Images for Ventricular Tachycardia Ablation Planning » Nicolas Cedilnik, Josselin Duchateau, Rémi Dubois, Frédéric Sacher, Pierre Jaïs, Hubert Cochet, Maxime Sermesant. EP-Europace, Oxford University Press, in press. Deep learning methods for image analysis were developed for the cardiac image analysis part, with participation to the STACOM 2018 challenge on atrial MRI segmentation: “Automatically Segmenting the Left Atrium from Cardiac Images Using Successive 3D U-Nets and a Contour Loss “, Shuman Jia, Antoine Despinasse, Zihao Wang, Hervé Delingette, Xavier Pennec, Pierre Jaïs, Hubert Cochet, and Maxime Sermesant.
4)Progress for the Sunnybrook team includes the development of a deformable registration method between cine MRI data and multi-contrast late enhancement (MCLE) images of scar. This method was developed to overcome issued related to the 2D slice misalignment of the acquired images (20 phases/heart cycle) between heart cycles due to motion, with an aim to generate accurate and smooth 3D anatomical heart models. A paper based on the results of this method was accepted and presented at the international cardiac workshop STACOM 2018 (at MICCAI conference, Granada, Spain), and will be published in Lecture Notes in Computer Science series by Springer: “Cine and Multicontrast Late Enhanced MRI Registration for 3D Heart Model Construction” Fumin Guo, Mengyuan Li, Matthew Ng, Graham Wright, Mihaela Pop.
2. Next year’s work program
Year 3: Assess the capacity of the personalised modelsto predict the critical sites for ablation therapyand guide the intervention.
We will validate the predictions of the developed models with the actual measures of abnormal heart rhythms and the results from the performed ablation therapy.
There is a pre-clinical protocol in Toronto that will start in 2019 in order to validate on in vivo porcine data the predictions with high resolution catheter mapping. The processing pipeline is now well established so these cases should be processed in the required time.
Image integration will be tested with the inHEART start-up company, in order to ensure a good 3D registration of the catheter data with the pre-operative imaging data, as this will be crucial for the accuracy of the validation data.
3. Record of activities
- Members of the Inria team (PI Maxime Sermesant and PhD student Nicolas Cedilnik) visited Sunnybrook team during the CMM Fields-Inria meeting in Toronto (January 2018). The teams worked together on building a framework to use mesh-less heart models built by Inria team for simulation of activation maps and intracardiac bipolar waves recorded in vivo from infarcted pig hearts by Sunnybrook team.
- 2. Sunnybrook PI (Mihaela Pop) visited Inria team in August 2018. The teams worked together on verifying all the reconstructed diffusion tensors for the fibre atlas, and planned next steps on developing statistical methods for studying changes in fibre directions in the border zone and collagenous scar, which will be important for electro-mechanical simulations of pathology.