Loading...

Upgrading the Ultrasound Imaging System Based on The Implementation of the Strain Imaging Mode

Fathi, Haniyeh | 2023

69 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56150 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Kavehvash, Zahra
  7. Abstract:
  8. Strain imaging is a non-invasive ultrasound modality for assessing cardiac function in echocardiography systems. In this thesis, we implemented a fully automated strain imaging system containing 5 steps: 1) echocardiographic view recognition, 2) cardiac cycle phase detection, i.e., the events of end-diastole (ED) and end-systole (ES), 3) segmentation of left ventricular (LV) myocardium, 4) motion estimation of this wall and 5) strain calculation. In this work, we propose a novel deep learning-based framework for phase detection of cardiac cycle by the use of echocardiographic images in multibeat videos. Further, by applying the augmentation technique, the model has been able to detect events in slow heart beats as well. Also, for the first time in this work we utilized the conditional generative adversarial network (cGAN) for LV segmentation task and we achieved appropriate results. In each of the mentioned stages of strain imaging, the results obtained are as follows: Classification of 5 main echocardiographic views has been done with 96.4% accuracy. Multibeat prediction of ED/ES events has an average absolute frame difference of 1.31 and 1.80 frames, respectively. The performance of the model in correctly detecting the number of heartbeats has been improved compared to previous methods. Dice coefficient for the task of LV segmentation is 94.85% which its performance is better than other methods in this field. Finally, we solve the motion estimation of LV myocardium with optical flow algorithms and then calculation of strain and strain curves has been done
  9. Keywords:
  10. Medical Ultrasound Imaging ; Strain Imaging ; Left Ventricular Segmentation ; Echocardiographic View Recognition ; Cardiac Phase Detection

 Digital Object List

 Bookmark

No TOC