Overview
This project focuses on leveraging Transformer-based machine learning models to predict cell trajectories in microfluidic devices. Accurate trajectory prediction is critical for optimizing the design of devices used in deterministic lateral displacement (DLD), which are widely used in cell sorting, rare cell isolation, and disease diagnostics.
The insights gained from this model can greatly enhance the design and functionality with a first VR MODEL 'RideV1' (first VR ride in Bangladesh) in the 'Bangladesh Smart city Expo and Conference, 2019', Xtreme Ride, and also intregration with developer team, a history-based educational Augmented Reality (AR) app on Language Movement of 21st February of 1952.
Key Functionalities
- Developed and implemented the design and implementation of the Transformer-based machine learning model for cell trajectory prediction
- Built and trained the Transformer model to capture complex spatiotemporal cell displacement in microfluidic environments
- Optimized the model into the optimization pipeline for DLD device designs, improving cell sorting accuracy and device efficiency
- Collaborated with biomedical researchers to validate the model's performance in real-world microfluidic systems
Current Progress
The project is in the model optimization phase, where the Transformer architecture has been built and tested on microfluidic datasets. The model shows promising results in accurately predicting cell trajectories, and the next phase involves further testing and integrating the model with other microfluidic design tools to enhance performance.
- Milestones: Model development, trained on microfluidic datasets
- Current Focus: Model optimization and integration with DLD device pipeline
- Future Goals: Expand applications to other microfluidic device designs beyond DLD