The Signal Science Lab is developing fundamental data processing methods that interface between biomedical techniques and biological systems to extract high temporal and spatial resolution information, enabling and aiding various spectroscopic, microscopic and imaging methods to study biological systems in desired environments.
To overcome noise problem in physical methods, we are developing wavelet transform-based noise reduction algorithms that include
- multi-dimensional representation of signals
- development of customized wavelets
- enhancement in signal resolution in the wavelet domain
- development of noise thresholds
Inverse Problems and Uncertainty Analysis
In many physical systems, the data acquired from the experiments is the indirect information that needs to be transformed to obtain the final solution. But there is no unique solution that exist from the data, hence ill-posed problem, and one must find accurate solution among several candidates. Current methods compromise resolution to avoid instability. We are developing a “localized” approach, where
- a mechanism is designed to optimize the solution at each output point independently
- uncertainty in results is quantified for for reliability
Protein Dynamics Studies using ESR Spectroscopy
Many biological studies focus on structure determination of soluble and membrane proteins but the study of dynamics is vital to understand disease mechanisms, including their function and interaction with their environment. Electron Spin Resonance (ESR) spectroscopy is a powerful method for studying structural dynamics of proteins at physiological temperatures for a wide range of time scales and can provide a detailed description of motion that includes both dynamics as well as local structural ordering. To address this problem of sensitivity in physiological environments, we are developing computational methods based on wavelet transforms to remove noise for accurate signal recovery.
Measuring Oxygen Concentration for Radiation Therapy
Oxygen concentration in tissue has proven to be an excellent marker for tumor composition and can be effectively used to determine radiation dose. Electron Spin Resonance Imaging (ESRI) can quantify pO2 and have shown to be effective in live animals. However, the challenge has been to translate ESRI for imaging humans in clinical settings because the dosage of paramagnetic probe for high quality imaging exceeds the human safety levels and needs to be reduced by an order-of-magnitude to make ESRI for humans feasible. The reduction in paramagnetic probe dose weakens the signal strength and increases the noise presence, lowering the signal-to-noise ratio (SNR) and affecting the pO2 measurement accuracy. The poor quality ESRI image obtained may also become clinically irrelevant. We aim to enable clinical ESRI studies by enhancing its sensitivity at desired paramagnetic probe concentrations through advanced techniques of signal processing and adapting wavelet denoising methods.
Reducing Scan Time for Clinical MRI
Magnetic Resonance Imaging (MRI) is widely used as a clinical diagnostic tool. Over 30 million MRI scans are conducted each year in the US. However, patients in need of MRI scans often face long wait-times as a result of imaging facility backlogs. The per-patient scan-time is a fundamental bottleneck that limits daily throughput, and is one of the root causes of backlogs and long wait-times. We are developing a signal-processing based software approach for reducing MRI scan-times. Our technology is focused on the novel application of denoising raw MRI data in real-time prior to construction of the MRI image and can be integrated with existing instrumentation without requiring any hardware modifications.