Designing an Intelligent System to Analyze Electrograms of Induced Pluripotent Stem Cell-Derived Cardiomyocytes

Golgooni, Zeinab | 2017

1879 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 50285 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rabiee, Hamid Reza; Soleymani, Mahdieh; Pahlavan, Sara
  7. Abstract:
  8. Ability to differentiate induced pluripotent stem cells to cardiomycocytes has attracted attentions,considering crucial role of the heart in the human body and great potential applications of these cells like disease modeling, new treatment methods and basic research. We are able to analyze the performance of beating cells through recording extracellular field potentials of cardiomyocytes using multi-electrode array (MEA) technology. This analysis is an essential step to use cardiac cells in any future development and experiment. Currently, the electrophysiology experts analyze recorded extracellular field potentials of induced cardiomyocytes by observing all the episodes of each record. This is an expensive, time-consuming, and subject-to-error task. To overcome this issue, we introduce a new, advanced, and automated method to analyze electrograms of iPSC-derived cardiomyocyte which can reduce cost, save electrophysiology experts time and improve accuracy of diagnosis.In this thesis, we propose a new classification method of iPSC-derived CMs with deep learning approach and use autmoated feature extraction phase instead of using hand-engineered features. We design a novel approach using two popular deep learning model, deep convolutional and recurrent neural networks. Our method can categorize new records to normal or arrhythmic by taking raw data without need to complex preprocessing or segmentation. In preparing step, a recording (with long duration) will be splited into smaller windows. After that, classification phase is made up of two main steps. In the first step, smaller windows are evaluated using the designed convolutional neural network. In the second step, the results of assessing smaller windows is given to the designed recurrent neural network as a sequence and output of this step will determine the label of main recording. We build a dataset of iPS cardiomyocyte extracellular field potentials recorded by the MEA technology in collaboration with the Royan stem cell institute. This dataset contains samples from both normal and abnormal category. This dataset has been used for training and testing the main method and comparative ones. In the experimantal results, we compare the result of the proposed method against methods based on the traditional machine learning. The results demonstrate that our mathed outperforms the two traditional methods. Our method has achieved the accuracy of 87%. In addition, the proposed method has the great ability to grow in the future; The performance will be enhanced by collecting new samples in the future and can be extended to the multi-class classification easily
  9. Keywords:
  10. Machine Learning ; Deep Neural Networks ; Data Analysis ; Multi-Electrode Arrays (MEA) ; Cardiomyocytes

 Digital Object List


...see more