Wellnest ECG
Wellnest takes your cardiac care to the next level with their proprietary Wellnest 12L ECG Machine which is equipped with advanced AI technology and a portable design.
TECHNOLOGY
TensorFlow, PyTorch, Python
PROJECT TYPE
Self-Supervised
Learning
Use Case
- Wellnest Health Monitoring Pvt. Ltd. wanted to train a custom AI model based on their ECG Dataset. The model would provide ECG Findings, Interpretations and Recommendations for the input ECG.
- The dataset consisted of raw bytes that were captured using their proprietary portable ECG device.
- The dataset was unlabelled and provided only basic information like gender and age of the patient.
- The raw bytes of the ECG were unfiltered and contained noise from surrounding electricity, electrodes, and patient's body.
Solution
- Dataset Generation
The raw ECG data set was cleaned of any noised to begin with. The dataset was then augmented in multiple ways to increase the number of dataset. It was then converted into images to be fed into the AI model.
- Self-supervised learning
Since the dataset was not labelled, we opted for self-supervised learning algorithms that included CNNs to train the AI model. Once trained the efficiency of the algorithm was measured through Silhouette score.
- Orchestrating AI Workflows with LangChain
We utilized LangChain to bring everything together by combining the output of the self-supervised learning model with the natural language processing to get the Findings, Interpretations, and Recommendations.