Research

​Introduction to the research in NICA lab

Acquiring Big Data from Brain

Imaging Brain Activity


​With genetic modification, the neurons can be modified to emit fluorescent light as a function of the brain activity (i.e., the neurons "blink" as they fire) which literally makes the brain activity visible. The main challenge is to record the optical signals at a high throughput and we develop and apply optical imaging techniques to tackle this.


Computational Imaging


Performance of imaging system is limited by many factors including the laws of physics, biological constraints, information theory and sampling theorem. We develop computational imaging methods that overcomes such limitation by predicting extensive information from the limited data by exploiting the prior knowledge of the distribution of the data.

Multiplexed Imaging


With fluorescence microscopy, only up to five proteins can be simultaneously imaged due to the wide emission spectra of fluorescent molecules. To image large number of proteins beyond this limit, we use spectrally overlapping fluorescent molecules to image large number of proteins and use signal processing algorithms to unmix images.


Analyzing Big Data from Brain

Neuro-image Processing


State-of-the-art functional imaging methods can generate more than a gigabyte of data per second which necessitates the development of automated analysis algorithms. We develop fast and scalable AI algorithms that can process brain images without any labeled data.

Neuro-data Mining


Neural activity is the basis of various operations in our brain, but our understanding of the fundamental principle of neural signal processing is very limited. We develop and apply computational methods to analyze the brain activity data to quantify how the information flows and reveal the functional connection among the neurons. Finding the repeating motifs, finding local circuits that operate together, and extracting the synaptic strength from the brain activity will lead us to deeper understanding.

Artificial Intelligence & Biomedical Engineering

Bio-inspired AI


One of the ultimate goals in the field of artificial intelligence is to design an AI that truly mimics a biological intelligence. We attempt to achieve this by bringing characteristics of biological intelligence, in terms of both architectures and learning process, to artificial intelligence.


Medical AI


Medical diagnosis is often based on medical images of patients that can be formulated as image classification problems which AI excels on. However, the variation in the imaging condition and equipment poses a major challenge in bringing AI to real life diagnosis. Moreover, only limited amount of labeled data is available in most scenarios. We overcome this limitation by combining AI techniques and understanding of the characteristics of the data.