AI-DRIVEN SIGNAL ENHANCEMENT OF LOW-RESOLUTION IMAGES

Smart overview of the Invention

The described technology utilizes AI to enhance low-resolution images, facilitating accurate sequencing of nucleic acid materials under such conditions. By employing a training set of images captured in both high and low-resolution scenarios, a model is trained to transform low-resolution images into enhanced versions resembling high-resolution captures. This approach aims to improve cost and operational efficiency in sequencing processes.

Background and Challenges

Current sequencing technologies, such as Next Generation Sequencing (NGS), handle massive volumes of data efficiently but face limitations due to imaging constraints and the density of clusters on flow cells. Optical instruments have intrinsic limits, like the Abbe diffraction limit, which restrict the density of nanowells on flow cells, affecting throughput and cost. While advanced optical components can enhance resolving power, they are expensive and do not completely overcome diffraction limitations.

AI-Driven Enhancements

The AI-driven systems described offer a method to process images at low resolving power, achieving high-resolution results without the need for costly optical upgrades. By training a model with paired high and low-resolution images, the system can apply trained filters to recover enhanced images from low-resolution inputs. This approach reduces the need for super-resolution methods like Structured Illumination Microscopy (SIM), which are computationally intensive and risk damaging nucleotide targets.

System Implementation

A computing device, such as an optical sequencing instrument, can implement this AI-driven enhancement. The device uses a processor to apply trained filters to low-resolution image data, producing enhanced resolution images suitable for base calling operations. This method simplifies the equipment required and reduces computational steps compared to traditional super-resolution techniques.

Advantages and Applications

The AI-driven enhancement supports increased nanowell density on flow cells, potentially up to three times greater than current technologies, without relying on SIM or high-numerical aperture objectives. This capability allows for higher throughput and more efficient sequencing operations. The technology offers practical solutions for imaging high-density flow cells, improving sequencing efficiency while minimizing costs and equipment complexity.