Real-time spectral analysis is often limited by high computational cost and algorithm complexity. Current computational techniques, using the Fourier Transform, for spectral estimation are not suitable for real time analysis, as each new time window must be recomputed. While it is possible to use the Fourier Transform for real-time spectral analysis, this requires significant computing power and sacrificing spatial resolution for speed. A team of Columbia researchers, led by Dr. Edward Ciaccio, has developed a new spectral estimator (NSE) algorithm for the real-time analysis of multichannel spectral imaging. The NSE takes the ensemble average of signal segments at multiple segment lengths. This approach ultimately expands the computational capacity while increasing analysis speed without sacrificing spatial resolution.
By using signal averages, the NSE is able to update and average signals using fewer computational operations. Using the NSE method, hundreds of data channels can be analyzed simultaneously without sacrificing the speed. Real-time NSE based analysis can potentially identify and characterize spectral transients, which often help physicians identify subtle diagnostic abnormalities. The NSE can be implemented as hardware or software to reduce computation and financial cost.
The increased efficiency of the NSE has been demonstrated using fractionated atrial electrogram data for catheter ablation of the atrial fibrillation substrate.
Patent Pending
Tech Ventures Reference: CU14084