We mostly work on Magnetic Resonance Imaging (MRI) and a bit on X-Ray Computed Tomography (CT). They are two very different imaging modalities and the challenges are different as well. But the underlying signal processing techniques are similar.

MRI is a versatile imaging modality. It is safe and can produce very quality images. But the main hindrance in its widespread application is its slow data acquisition time. So how do we tackle this?

The only way to speed up the scan time is to acquire less data. Now the question arises - how do we reconstruct the image from less data? Compressed Sensing (CS) has played an integral part in answering this questions in the last half a decade. CS reconstructs MR images by exploiting their spatial redundancies (for static MRI) or spatio-temporal redundancies (for dynamic MRI).

We work on almost all the different MRI problems; but most notably on multi-channel parallel MRI and dynamic MRI. We have devised robust techniques for parallel MRI with a fitting name - CaLM (Calibration Less Multi-channel) MRI. It performs better than commercial (GRAPPA) and state-of-the-art (l1 SPIRiT) methods; but its major advantage stems from the fact that it does not require any calibration or parameter estimation.

l1 SPIRiT

GRAPPA

CaLM

In dynamic MRI, the challenge is to reduce the scan time in order to improve temporal resolution. The best reconstruction methods fuse multiple approaches of sparse and low-rank recovery to achieve this goal. We too have some results worth noticing.

In CT imaging, the problem is different. CT is harmful owing to its ionizing radiation. How can we make it safer? - We got to reduce the subject's exposure to ionizing radiation without compromising on image quality. The only way this can be achieved is by reducing the amount of data we usually acquire. Again CS comes to the aid. The approach here is similar to MRI - even though the goals are different.