# Detection and segmentation of the Left Ventricle in Cardiac MRI using Deep Learning

**Authors:** Alexandre Attia, Sharone Dayan  
**arXiv:** [1801.02171](https://arxiv.org/abs/1801.02171)  
**Date:** 2018-01-07  
**Summary page:** https://www.summarizepaper.com/en/paper/1801.02171v1/

## Key points

- Manual segmentation of the Left Ventricle (LV) in cardiac diagnostics is a laborious and meticulous task
- Automation of this process is needed to improve accuracy and efficiency
- Avendi et al. proposed using Convolutional Neural Networks (CNNs), Stacked Auto-Encoders, and Deformable Models for automated segmentation
- Experiments were conducted to evaluate the effectiveness of their proposed method
- Results showed that deep learning techniques can achieve more accurate segmentations compared to manual delineation by experts
- Avendi et al.'s approach allows for experimentation with different variations, leading to potential advancements in automated LV segmentation.

## Summary

In the field of cardiac diagnostics, manual segmentation of the Left Ventricle (LV) is a laborious and meticulous task that can vary depending on factors such as the patient, Magnetic Resonance Images (MRI) cuts, and the expertise of the individuals involved. Currently, manual delineation performed by experts is considered to be the ground truth for cardiac diagnosticians. However, there is a need to automate this process in order to improve accuracy and efficiency. To address this challenge, Avendi et al. proposed utilizing Convolutional Neural Networks (CNNs), Stacked Auto-Encoders, and Deformable Models to automate the segmentation process while achieving higher levels of accuracy. In order to evaluate the effectiveness of their proposed method, parts of Avendi et al. 's paper were implemented and experiments were conducted. Both the original method described in the paper and slightly modified versions with changes made to the architecture and parameters were tested. The results obtained from these experiments provide valuable insights into the potential of deep learning techniques for automating LV segmentation in cardiac MRI. By combining CNNs, Stacked Auto-Encoders, and Deformable Models it is possible to achieve more accurate segmentations compared to manual delineation by experts. This refined summary highlights not only the challenges associated with manual LV segmentation but also introduces Avendi et al. 's innovative approach using deep learning techniques which emphasizes how implementing parts of their proposed method allowed for experimentation with different variations leading to improved understanding and potential advancements in automated LV segmentation.