Higher resolution MRI acquired faster
TLDR; MICSI is introducing an innovative image processing algorithm to decrease MRI scan time, increase image resolution, and unlock new diagnostic capabilities using Microstructure Imaging.
Problem: MRI Centers are overbooked.
If you've ever had an MRI, you've probably waited weeks for an available slot and then spent a long time in the scanner. These long scan times inconvenience patients, create hospital backlogs, and delay appointments. Delays like these can critically impact conditions that require prompt diagnosis and intervention, inflate healthcare costs, increase patient anxiety, and potentially compromise treatment strategies. Why the wait? An MRI of the whole brain can be acquired in seconds!
Compromise: Why is your MRI an hour long?
Most of the scan time is spent on preserving signal-to-noise ratio (SNR) to produce images of the highest quality. These standard SNR preserving methods include:
An additional compromise is related to the spatial resolution, since MRI is acquired in 3D with “voxels”, doubling the resolution in each dimension from (2x2x2 mm3 to 1x1x1 mm3) would require the SNR to increase by 8x, which even after 50 years of innovation, seems near impossible.
The MRI exam is full of compromise.
We believe that the ultimate compromise is that MRI is not being used to its fullest potential for microstructure imaging, where MRI could be used to image cellular properties non-invasively.
Solution: Boosting SNR over the MRI exam using MICSI-RMT.
Our patented approach (pending FDA510k clearance in Q4-2023) uses a self-supervised AI method to learn the noise properties of the MRI exam and remove noise uniformly across all images of the dataset. Our algorithm, MICSI-RMT, can be described as a smart averaging approach, whereby many images of the MRI exam are combined while preserving their unique image properties and discarding the noise. This boost in SNR would be approximately, sqrt(M/P), where M is the number of images included in the dataset, whereas P describes how different the images are (physical properties: proton density, T1, T2, diffusion, etc + artifacts: motion, errors, etc).
MICSI-RMT is the commercial implementation of the MP-PCA algorithm that was developed by our research group led by Drs. Dmitry S. Novikov and Els Fieremans of NYU Radiology’s Center for Biomedical Imaging.
MP-PCA is the most popular denoising approach in the MRI community:
With MICSI-RMT, MRI centers will not only be able to provide higher quality imaging but also drastically reduce the scan time (up to 50%). This improved efficiency allows centers to scan more patients per day, enhancing their operational capacity and throughput. Financially, this increase in capacity translates into a significant boost in revenue for every MRI center. By doubling the number of patients scanned daily, we estimate an additional $2 million in annual revenue per MRI scanner. Furthermore, the improved image quality and resolution could lead to more accurate diagnoses, early disease detection, and reducing overall healthcare costs in the long run.
Who we are - Greg and Ben
We are scientists and engineers who have spent the last decade developing machine learning software to both improve the quality diagnostic utility of MRI. Gregory Lemberskiy, PhD is an experimental physicist. He discovered how to use MRI to measure the diameter of the prostate glandular lumen, which may obviate the need for tissue biopsy to detect prostate cancer progression. Benjamin Ades-Aron, PhD is an electrical engineer who developed and deployed the DESIGNER open-source image processing pipeline, which is the global gold standard of diffusion MRI image processing. He has also developed a number of novel neural network architectures to aid in the denoising of compressed sensing MRI data, to facilitate brain mapping for functional neurosurgery of Epilepsy and brain tumor.
Together, the MICSI team owns over a dozen approved patents and has a clear vision for the future of MRI. The goal of MICSI is to use our patented denoising and parameter estimation technologies to make MRI quantitative and reproducible for the first time, creating a new category in diagnostic healthcare and improving patient satisfaction and outcomes.