Open access image repositories: high-quality data to enable machine learning research

Originally motivated by the need for research reproducibility and data reuse, large-scale, open access information repositories have become key resources for training and testing of advanced machine learning applications in biomedical and clinical research. To be of value, such repositories must provide large, high-quality data sets, where quality is defined as minimising variance due to data collection protocols and data misrepresentations. Curation is the key to quality. We have constructed a large public access image repository, The Cancer Imaging Archive, dedicated to the promotion of open science to advance the global effort to diagnose and treat cancer. Drawing on this experience and our experience in applying machine learning techniques to the analysis of radiology and pathology image data, we will review the requirements placed on such information repositories by state-of-the-art machine learning applications and how these requirements can be met.

An Assessment of Imaging Informatics for Precision Medicine in Cancer

Objectives: Precision medicine requires the measurement, quantification, and cataloging of medical characteristics to identify the most effective medical intervention. However, the amount of available data exceeds our current capacity to extract meaningful information. We examine the informatics needs to achieve precision medicine from the perspective of quantitative imaging and oncology. Methods: The National Cancer Institute (NCI) organized several workshops on the topic of medical imaging and precision medicine. The observations and recommendations are summarized herein. Results: Recommendations include: use of standards in data collection and clinical correlates to promote interoperability; data sharing and validation of imaging tools; clinician's feedback in all phases of research and development; use of open-source architecture to encourage reproducibility and reusability; use of challenges which simulate real-world situations to incentivize innovation; partnership with industry to facilitate commercialization; and education in academic communities regarding the challenges involved with translation of technology from the research domain to clinical utility and the benefits of doing so. Conclusions: This article provides a survey of the role and priorities for imaging informatics to help advance quantitative imaging in the era of precision medicine. While these recommendations were drawn from oncology, they are relevant and applicable to other clinical domains where imaging aids precision medicine.

The Human Connectome Project: a data acquisition perspective

The Human Connectome Project (HCP) is an ambitious 5-year effort to characterize brain connectivity and function and their variability in healthy adults. This review summarizes the data acquisition plans being implemented by a consortium of HCP investigators who will study a population of 1200 subjects (twins and their non-twin siblings) using multiple imaging modalities along with extensive behavioral and genetic data. The imaging modalities will include diffusion imaging (dMRI), resting-state fMRI (R-fMRI), task-evoked fMRI (T-fMRI), T1- and T2-weighted MRI for structural and myelin mapping, plus combined magnetoencephalography and electroencephalography (MEG/EEG). Given the importance of obtaining the best possible data quality, we discuss the efforts underway during the first two years of the grant (Phase I) to refine and optimize many aspects of HCP data acquisition, including a new 7T scanner, a customized 3T scanner, and improved MR pulse sequences.

Alcoholic fixation of blood to surgical instruments—a possible factor in the surgical transmission of CJD?

While developing a new protein removal test for the quality control of surgical instrument cleaning, it was noted that alcohol firmly binds blood to stainless steel. Creutzfeldt-Jacob disease is one of the transmissible spongiform encephalopathies (TSE) that has been transmitted between humans and chimpanzees by electroencephalogram electrodes, previously 'sterilized' using ethanol and formaldehyde. Although ethanol has a bactericidal action, it also binds protein to metal. Prion proteins found in TSE are thought to be the causal agents of spongiform disease and it is likely that these proteins are also bound to the stainless steel of surgical instruments by alcohols. Where spongiform disease is a possibility, alcohol, and probably formaldehyde, should not be used to decontaminate neurosurgical instruments.

Sorting of glycoprotein B from human cytomegalovirus to protein storage vesicles in seeds of transgenic tobacco

As part of ongoing studies into the use of plant expression systems for making human therapeutic proteins, we have successfully expressed the major glycoprotein, gB, of human cytomegalovirus (HCMV) in transgenic tobacco plants. Viral glycoprotein was detectable in the protein extracts of mature tobacco seeds using neutralizing and non-neutralizing monoclonal antibodies specific for gB. Although several mammalian proteins have been expressed in tobacco, localization of these proteins in transgenic tobacco tissue has not been extensively examined. The objective of this study was to identify the site(s) of recombinant gB deposition in mature tobacco seeds. Using immunogold labelling and electron microscopy, we found specific labelling for gB in the endosperm of transgenic seeds, with gB localized almost exclusively in protein storage vesicles (PSV). This occurred in seeds that were freshly harvested and in seeds that had been stored for several months. These data indicate that gB behaves like a plant storage protein when expressed in tobacco seeds, and provide further support for the suitability of plants for producing recombinant proteins of potential clinical relevance.

Development of biopharmaceuticals in plant expression systems: cloning, expression and immunological reactivity of human cytomegalovirus glycoprotein B (UL55) in seeds of transgenic tobacco

Plant seeds offer unique opportunities for the production and delivery of oral subunit vaccines. We have used the immunodominant glycoprotein B complex of human cytomegalovirus (HCMV), introduced into tobacco plants, as a model system for studying the merit of this promising approach. Given the advantages of expressing proteins in seeds, a novel expression vector was developed incorporating regulatory sequences of glutelin, the major rice seed storage protein, to direct synthesis of recombinant glycoprotein B. Analysis of genomic DNA of 28 selected tobacco transformants by PCR amplification showed that 71% harboured the gB cDNA, a finding further documented by Southern blotting. Specific immunoassays of protein extracts from seeds of positive plants showed that all were producing antigenic glycoprotein B at levels ranging from 70-146 ng/mg extracted protein. In addition, similarity with native glycoprotein B produced in HCMV-infected cells was also demonstrated by inhibition of immunofluorescence on HCMV-infected human fibroblasts. These data are the first to report the expression of an immunodominant antigen of HCMV in plant tissues, indicating the fidelity with which this very large heterologous viral glycoprotein can be synthesized in this model system.

The role of digital imaging and communications in medicine in an evolving healthcare computing environment: the model is the message

The decision to use Digital Imaging and Communications in Medicine (DICOM), Health Level 7 (HL7), a common object broker such as the Common Object Request Brokering Architecture (CORBA) or ActiveX (Microsoft Corp, Redmond, WA) or any other protocol for the transfer of DICOM data depends on the requirements of a particular implementation. The selection of protocol is independent of the information model. Our goal as message standards developers is to design a data interchange infrastructure that will faithfully convey the computer-based patient record and make it available to authorized health care providers when and where it is needed for patient care. DICOM accurately and expressively represents the clinically significant properties of images and the semantics of image-related information. The DICOM data model is small and well-defined. The model can be expressed in Standard Generalized Markup Language (SGML) or Object Management Group Interface Definition Language or other common syntax-and can be implemented using any reliable communications protocol. Therefore our opinion is that the DICOM semantic data model should serve as the basis for a logically equivalent set of specifications in HL7, CORBA, ActiveX, and SGML for the interchange of biomedical images and image-related information.