IDEAL: Images Across Domains, Experiments, Algorithms and Learning

Abstract

Research across science domains is increasingly reliant on image-centric data. Software tools are in high demand to uncover relevant, but hidden, information in digital images, such as those coming from faster next generation high-throughput imaging platforms. The challenge is to analyze the data torrent generated by the advanced instruments efficiently, and provide insights such as measurements for decision-making. In this paper, we overview work performed by an interdisciplinary team of computational and materials scientists, aimed at designing software applications and coordinating research efforts connecting (1) emerging algorithms for dealing with large and complex datasets; (2) data analysis methods with emphasis in pattern recognition and machine learning; and (3) advances in evolving computer architectures. Engineering tools around these efforts accelerate the analyses of image-based recordings, improve reusability and reproducibility, scale scientific procedures by reducing time between experiments, increase efficiency, and open opportunities for more users of the imaging facilities. This paper describes our algorithms and software tools, showing results across image scales, demonstrating how our framework plays a role in improving image understanding for quality control of existent materials and discovery of new compounds.

Fracture mechanical characterization of the initiation and growth of interior fatigue cracks

“…The second problem is that LEFM principles were primarily developed for long cracks, and not for small cracks. Various other strategies have been developed recently …”

Radial feature descriptors for cell classification and recommendation

“…The combined use of shape and texture features was also reported by Mariarputham et al [9], who used 7 features sets that included the relative size of the nucleus and cytoplasm, the dynamic range, and the first 4 moments of the intensities, relative displacement of nucleus within the cytoplasm, Gray Level Cooccurrence Matrix (GLCM) [13] Previous investigations show that textural information plays a key role in the description of biological data, with applications ranging from face recognition [33,34,35] to cell classification [4,36,37] of cervical cells from Pap smears, which motivates our proposal of a new descriptor based on the nucleus edge; it computes the texture information around the nuclear membrane using images from the Herlev and CRIC databases. Despite the presence of many different scientific publications, they lack a common metric to compare the cell analysis performance, therefore we propose the evaluation of 14 different cell description methods using the same classifier and the same performance metrics: false negative rate and the κ index.…”

“…A Content-Based Image Retrieval (CBIR) system is a search engine that uses similarity metrics to retrieve and rank images from a database by comparing their feature vectors to an input query [36,37]. Besides the classification procedure, we also performed the CBIR experiments to explore the RFD generalization for image recommendation tasks.…”

Convolutional neural networks at the interface of physical and digital data

“…Figure 6(b) emphasizes several fiber profiles, commonly used as a guiding pattern to detect other fiber cross-sections. A major demand is the ability to perform pattern ranking mechanisms, which can steer algorithms such as template matching [12,13], and improve data management by beamline scientists.…”