Point-of-care cervical cancer screening using deep learning-based microholography

Most deaths (80%) from cervical cancer occur in regions lacking adequate screening infrastructures or ready access to them. In contrast, most developed countries now embrace human papillomavirus (HPV) analyses as standalone screening; this transition threatens to further widen the resource gap.

Methods: We describe the development of a DNA-focused digital microholography platform for point-of-care HPV screening, with automated readouts driven by customized deep-learning algorithms. In the presence of high-risk HPV 16 or 18 DNA, microbeads were designed to bind the DNA targets and form microbead dimers. The resulting holographic signature of the microbeads was recorded and analyzed.

Results: The HPV DNA assay showed excellent sensitivity (down to a single cell) and specificity (100% concordance) in detecting HPV 16 and 18 DNA from cell lines. Our deep learning approach was 120-folder faster than the traditional reconstruction method and completed the analysis in < 2 min using a single CPU. In a blinded clinical study using patient cervical brushings, we successfully benchmarked our platform's performance to an FDA-approved HPV assay.

Conclusions: Reliable and decentralized HPV testing will facilitate cataloguing the high-risk HPV landscape in underserved populations, revealing HPV coverage gaps in existing vaccination strategies and informing future iterations.

Design and clinical validation of a point-of-care device for the diagnosis of lymphoma via contrast-enhanced microholography and machine learning

The identification of patients with aggressive cancer who require immediate therapy is a health challenge in low-income and middle-income countries. Limited pathology resources, high healthcare costs and large-case loads call for the development of advanced standalone diagnostics. Here, we report and validate an automated, low-cost point-of-care device for the molecular diagnosis of aggressive lymphomas. The device uses contrast-enhanced microholography and a deep-learning algorithm to directly analyse percutaneously obtained fine-needle aspirates. We show the feasibility and high accuracy of the device in cells, as well as the prospective validation of the results in 40 patients clinically referred for image-guided aspiration of nodal mass lesions suspicious for lymphoma. Automated analysis of human samples with the portable device should allow for the accurate classification of patients with benign and malignant adenopathy.

The global cardiovascular magnetic resonance registry (GCMR) of the society for cardiovascular magnetic resonance (SCMR): its goals, rationale, data infrastructure, and current developments

BACKGROUND

With multifaceted imaging capabilities, cardiovascular magnetic resonance (CMR) is playing a progressively increasing role in the management of various cardiac conditions. A global registry that harmonizes data from international centers, with participation policies that aim to be open and inclusive of all CMR programs, can support future evidence-based growth in CMR.

METHODS

The Global CMR Registry (GCMR) was established in 2013 under the auspices of the Society for Cardiovascular Magnetic Resonance (SCMR). The GCMR team has developed a web-based data infrastructure, data use policy and participation agreement, data-harmonizing methods, and site-training tools based on results from an international survey of CMR programs.

RESULTS

At present, 17 CMR programs have established a legal agreement to participate in GCMR, amongst them 10 have contributed CMR data, totaling 62,456 studies. There is currently a predominance of CMR centers with more than 10 years of experience (65%), and the majority are located in the United States (63%). The most common clinical indications for CMR have included assessment of cardiomyopathy (21%), myocardial viability (16%), stress CMR perfusion for chest pain syndromes (16%), and evaluation of etiology of arrhythmias or planning of electrophysiological studies (15%) with assessment of cardiomyopathy representing the most rapidly growing indication in the past decade. Most CMR studies involved the use of gadolinium-based contrast media (95%).

CONCLUSIONS

We present the goals, mission and vision, infrastructure, preliminary results, and challenges of the GCMR.

TRIAL REGISTRATION

Identification number on ClinicalTrials.gov: NCT02806193 . Registered 17 June 2016.

Holographic Assessment of Lymphoma Tissue (HALT) for Global Oncology Field Applications

Low-cost, rapid and accurate detection technologies are key requisites to cope with the growing global cancer challenges. The need is particularly pronounced in resource-limited settings where treatment opportunities are often missed due to the absence of timely diagnoses. We herein describe a Holographic Assessment of Lymphoma Tissue (HALT) system that adopts a smartphone as the basis for molecular cancer diagnostics. The system detects malignant lymphoma cells labeled with marker-specific microbeads that produce unique holographic signatures. Importantly, we optimized HALT to detect lymphomas in fine-needle aspirates from superficial lymph nodes, procedures that align with the minimally invasive biopsy needs of resource-constrained regions. We equipped the platform to directly address the practical needs of employing novel technologies for "real world" use. The HALT assay generated readouts in <1.5 h and demonstrated good agreement with standard cytology and surgical pathology.

Sparsity-Based Pixel Super Resolution for Lens-Free Digital In-line Holography

Lens-free digital in-line holography (LDIH) is a promising technology for portable, wide field-of-view imaging. Its resolution, however, is limited by the inherent pixel size of an imaging device. Here we present a new computational approach to achieve sub-pixel resolution for LDIH. The developed method is a sparsity-based reconstruction with the capability to handle the non-linear nature of LDIH. We systematically characterized the algorithm through simulation and LDIH imaging studies. The method achieved the spatial resolution down to one-third of the pixel size, while requiring only single-frame imaging without any hardware modifications. This new approach can be used as a general framework to enhance the resolution in nonlinear holographic systems.

Digital diffraction analysis enables low-cost molecular diagnostics on a smartphone

The widespread distribution of smartphones, with their integrated sensors and communication capabilities, makes them an ideal platform for point-of-care (POC) diagnosis, especially in resource-limited settings. Molecular diagnostics, however, have been difficult to implement in smartphones. We herein report a diffraction-based approach that enables molecular and cellular diagnostics. The D3 (digital diffraction diagnosis) system uses microbeads to generate unique diffraction patterns which can be acquired by smartphones and processed by a remote server. We applied the D3 platform to screen for precancerous or cancerous cells in cervical specimens and to detect human papillomavirus (HPV) DNA. The D3 assay generated readouts within 45 min and showed excellent agreement with gold-standard pathology or HPV testing, respectively. This approach could have favorable global health applications where medical access is limited or when pathology bottlenecks challenge prompt diagnostic readouts.

High throughput transmission optical projection tomography using low cost graphics processing unit

We implement the use of a graphics processing unit (GPU) in order to achieve real time data processing for high-throughput transmission optical projection tomography imaging. By implementing the GPU we have obtained a 300 fold performance enhancement in comparison to a CPU workstation implementation. This enables to obtain on-the-fly reconstructions enabling for high throughput imaging.

Enhancing navigation in biomedical databases by community voting and database-driven text classification

Background

The breadth of biological databases and their information content continues to increase exponentially. Unfortunately, our ability to query such sources is still often suboptimal. Here, we introduce and apply community voting, database-driven text classification, and visual aids as a means to incorporate distributed expert knowledge, to automatically classify database entries and to efficiently retrieve them.

Results

Using a previously developed peptide database as an example, we compared several machine learning algorithms in their ability to classify abstracts of published literature results into categories relevant to peptide research, such as related or not related to cancer, angiogenesis, molecular imaging, etc. Ensembles of bagged decision trees met the requirements of our application best. No other algorithm consistently performed better in comparative testing. Moreover, we show that the algorithm produces meaningful class probability estimates, which can be used to visualize the confidence of automatic classification during the retrieval process. To allow viewing long lists of search results enriched by automatic classifications, we added a dynamic heat map to the web interface. We take advantage of community knowledge by enabling users to cast votes in Web 2.0 style in order to correct automated classification errors, which triggers reclassification of all entries. We used a novel framework in which the database "drives" the entire vote aggregation and reclassification process to increase speed while conserving computational resources and keeping the method scalable. In our experiments, we simulate community voting by adding various levels of noise to nearly perfectly labelled instances, and show that, under such conditions, classification can be improved significantly.

Conclusion

Using PepBank as a model database, we show how to build a classification-aided retrieval system that gathers training data from the community, is completely controlled by the database, scales well with concurrent change events, and can be adapted to add text classification capability to other biomedical databases.

The system can be accessed at .

Born normalization for fluorescence optical projection tomography for whole heart imaging

Optical projection tomography is a three-dimensional imaging technique that has been recently introduced as an imaging tool primarily in developmental biology and gene expression studies. The technique renders biological sample optically transparent by first dehydrating them and then placing in a mixture of benzyl alcohol and benzyl benzoate in a 2:1 ratio (BABB or Murray s Clear solution). The technique renders biological samples optically transparent by first dehydrating them in graded ethanol solutions then placing them in a mixture of benzyl alcohol and benzyl benzoate in a 2:1 ratio (BABB or Murray s Clear solution) to clear. After the clearing process the scattering contribution in the sample can be greatly reduced and made almost negligible while the absorption contribution cannot be eliminated completely. When trying to reconstruct the fluorescence distribution within the sample under investigation, this contribution affects the reconstructions and leads, inevitably, to image artifacts and quantification errors.. While absorption could be reduced further with a permanence of weeks or months in the clearing media, this will lead to progressive loss of fluorescence and to an unrealistically long sample processing time. This is true when reconstructing both exogenous contrast agents (molecular contrast agents) as well as endogenous contrast (e.g. reconstructions of genetically expressed fluorescent proteins).

PepBank--a database of peptides based on sequence text mining and public peptide data sources

Background

Peptides are important molecules with diverse biological functions and biomedical uses. To date, there does not exist a single, searchable archive for peptide sequences or associated biological data. Rather, peptide sequences still have to be mined from abstracts and full-length articles, and/or obtained from the fragmented public sources.

Description

We have constructed a new database (PepBank), which at the time of writing contains a total of 19,792 individual peptide entries. The database has a web-based user interface with a simple, Google-like search function, advanced text search, and BLAST and Smith-Waterman search capabilities. The major source of peptide sequence data comes from text mining of MEDLINE abstracts. Another component of the database is the peptide sequence data from public sources (ASPD and UniProt). An additional, smaller part of the database is manually curated from sets of full text articles and text mining results. We show the utility of the database in different examples of affinity ligand discovery.

Conclusion

We have created and maintain a database of peptide sequences. The database has biological and medical applications, for example, to predict the binding partners of biologically interesting peptides, to develop peptide based therapeutic or diagnostic agents, or to predict molecular targets or binding specificities of peptides resulting from phage display selection. The database is freely available on , and the text mining source code (Peptide::Pubmed) is freely available above as well as on CPAN ().

MIPortal: A High Capacity Server for Molecular Imaging Research

The introduction of novel molecular tools in research and clinical medicine has created a need for more refined information management systems. This article describes the design and implementation of such a new information platform: the Molecular Imaging Portal (MIPortal). The platform was created to organize, archive, and rapidly retrieve large datasets using Web-based browsers as access points. The system has been implemented in a heterogeneous, academic research environment serving Macintosh, Unix, and Microsoft Windows clients and has been shown to be extraordinarily robust and versatile. In addition, it has served as a useful tool for clinical trials and collaborative multi-institutional small-animal imaging research.

Focal hepatic lesions: effect of three-dimensional wavelet compression on detection at CT

PURPOSE

To evaluate the effect of three-dimensional, wavelet-based compression on the detection of focal hepatic lesions at computed tomography (CT).

MATERIALS AND METHODS

CT images obtained in 69 patients with focal hepatic lesions were studied (35 consecutive cases and 34 cases selected to be difficult on the basis of lesion size or contrast). Image data were compressed by means of a three-dimensional, wavelet-based algorIthm at ratios of 10:1, 15:1, and 20:1. Normal and abnormal sections (on original and compressed images) were reviewed by using an interactive workstation. Four readers rated the presence or absence of a lesion with a five-point scale.

RESULTS

At receiver operating characteristic analysis, no statistically significant difference was detected for all cases considered together. Differences approached but did not reach statistical significance for the diagnostic performance of one reader with compressed images (15:1, P = .054; 20:1, P = .051). For the subset of difficult cases, a significant difference was observed with 20:1 compressed images for one reader (P = .026). Diagnostic performance of readers was less certain in normal than in abnormal cases with both original and compressed images (difference was significant for 15:1 [P = .035] and 20:1 [P < .0001] compressed images).

CONCLUSION

Three-dimensional wavelet compression is satisfactory at ratios of at least 10:1. Additional studies with a larger sample would help confirm findings with higher ratios.

Application of wavelet compression to digitized radiographs

OBJECTIVE

Image data compression is an enabling technology for teleradiology and picture archive and communication systems. Compression decreases the time and cost of image transmission and the requirements for image storage. Wavelets, discovered in 1987, constitute a new compression technique that has been described in radiologic publications but, to our knowledge, no previous studies of its use have been reported. The purpose of this study was to demonstrate the application of wavelet-based compression technology to digitized radiographs.

MATERIALS AND METHODS

Twelve radiographs with abnormal findings were digitized, compressed, and decompressed by using a new wavelet-based lossy compression algorithm. Images were compressed at ratios from 10:1 to 60:1. Seven board-certified radiologists reviewed images on a two-headed, high-resolution (2K x 2K) diagnostic workstation. Paired original and compressed/decompressed images were presented in random order. Reviewers adjusted contrast and magnification to judge whether image degradation was present, and if so, whether it was of diagnostic significance. Quantitative error measures were tabulated.

RESULTS

Reviewers found no clinically relevant degradation below a compression ratio of 30:1. Skeletal radiographs appeared more sensitive to compression than did chest or abdominal radiographs, but the trend was not statistically significant. Quantitative error measures increased gradually with compression ratio.

CONCLUSION

On the basis of subjective assessment of image quality and the computational efficiency of the algorithm, wavelet-base techniques appear promising for the compression of digitized radiographs. The results of this initial experience can be used to design appropriate observer performance studies.