Unlike conventional approaches where scientists develop independent, stand-alone imaging techniques to solve specific problems, STROBE will advance and integrate different imaging techniques by using common underpinning technologies - i) advanced algorithms; ii) fast detectors and data acquisition methods; iii) big data handling, analysis and visualization; and iv) adaptive multimodal imaging - to collectively address major science challenges. We will image same material or biological systems at multi-length and multi-dimensions using electron, x-ray and optical probe. The correlative approach will allow us to address important scientific questions that would otherwise be inaccessible by other means.
IV.1) Advanced algorithms for electron, x-ray and adaptive optical nanoimaging. One of the great strengths of STROBE is advanced algorithms and computational imaging, including the pioneering of nonlinear variational methods, CDI, atomic electron tomography (AET) and computational microscopy, all of which rely on powerful algorithms. Building upon our extensive expertise and successful previous collaborations, we will transform and accelerate imaging science by integrating algorithm development and image reconstruction activities across electron, x-ray and optical imaging modalities. This is in contrast to current approaches, which have historically been developed independently by each imaging field.
IV.2) Fast detectors and data acquisition methods for real-time functional imaging. The detector development has greatly contributed to the recent revolution in electron, x-ray and optical imaging. LBL, a STROBE partner laboratory, has played a key role in the development of the K2 direct detection detectors, FAST CCD detector and CMOS pixel sensor for electron microscopy and x-ray imaging. These detectors bring high-quality, low noise, quantitative x-ray imaging and electron microscopy at frame rates >100 Hz. By combining these fast direct detectors with real-time GPU-based data analysis, we will develop 4D functional microscopes that capture and display reconstructed images in real-time, while also adding new capabilities such as elemental/chemical identification of functioning systems. To manage large data, we will design computational methods for each imaging modality to enable efficient and fast data acquisition and processing.
IV.3) Big data handling, analysis and visualization. STROBE will implement real-time image reconstruction and visualization for all imaging modalities through a combination of GPU processing, parallel computing and more efficient algorithms. The dramatic increase in data volume generated by multi-dimensional and multi-scale functional imaging in STROBE calls for fast detectors and high performance reconstruction algorithms. STROBE will develop infrastructure for both data reduction on the collection side, as well as data management and algorithm efficiency on the data analysis side. Furthermore, we will implement efficient and parallelized versions of our algorithms. GPU programming with NVIDIA CUDA will enable 10-300x speedup on smaller (<12 Gigabytes) datasets for CDI and related techniques that can be reconstructed on-the-fly, while cluster-based computing can be used to process larger data.