Bridging the gap between prostate radiology and pathology through machine learning.

BACKGROUND: Prostate cancer remains the second deadliest cancer for American men despite clinical advancements. Currently, magnetic resonance imaging (MRI) is considered the most sensitive non-invasive imaging modality that enables visualization, detection, and localization of prostate cancer, and is increasingly used to guide targeted biopsies for prostate cancer diagnosis. However, its utility remains limited due to high rates of false positives and false negatives as well as low inter-reader agreements. PURPOSE: Machine learning methods to detect and localize cancer on prostate MRI can help standardize radiologist interpretations. However, existing machine learning methods vary not only in model architecture, but also in the ground truth labeling strategies used for model training. We compare different labeling strategies and the effects they have on the performance of different machine learning models for prostate cancer detection on MRI. METHODS: Four different deep learning models (SPCNet, U-Net, branched U-Net, and DeepLabv3+) were trained to detect prostate cancer on MRI using 75 patients with radical prostatectomy, and evaluated using 40 patients with radical prostatectomy and 275 patients with targeted biopsy. Each deep learning model was trained with four different label types: pathology-confirmed radiologist labels, pathologist labels on whole-mount histopathology images, and lesion-level and pixel-level digital pathologist labels (previously validated deep learning algorithm on histopathology images to predict pixel-level Gleason patterns) on whole-mount histopathology images. The pathologist and digital pathologist labels (collectively referred to as pathology labels) were mapped onto pre-operative MRI using an automated MRI-histopathology registration platform. RESULTS: Radiologist labels missed cancers (ROC-AUC: 0.75-0.84), had lower lesion volumes (~68% of pathology lesions), and lower Dice overlaps (0.24-0.28) when compared with pathology labels. Consequently, machine learning models trained with radiologist labels also showed inferior performance compared to models trained with pathology labels. Digital pathologist labels showed high concordance with pathologist labels of cancer (lesion ROC-AUC: 0.97-1, lesion Dice: 0.75-0.93). Machine learning models trained with digital pathologist labels had the highest lesion detection rates in the radical prostatectomy cohort (aggressive lesion ROC-AUC: 0.91-0.94), and had generalizable and comparable performance to pathologist label-trained-models in the targeted biopsy cohort (aggressive lesion ROC-AUC: 0.87-0.88), irrespective of the deep learning architecture. Moreover, machine learning models trained with pixel-level digital pathologist labels were able to selectively identify aggressive and indolent cancer components in mixed lesions on MRI, which is not possible with any human-annotated label type. CONCLUSIONS: Machine learning models for prostate MRI interpretation that are trained with digital pathologist labels showed higher or comparable performance with pathologist label-trained models in both radical prostatectomy and targeted biopsy cohort. Digital pathologist labels can reduce challenges associated with human annotations, including labor, time, inter- and intra-reader variability, and can help bridge the gap between prostate radiology and pathology by enabling the training of reliable machine learning models to detect and localize prostate cancer on MRI.

Management of urinary incontinence in residential care.

BACKGROUND: Urinary incontinence is prevalent in residential care and rates are expected to increase with the ageing population in Australia. It contributes to poor quality of life (QoL), functional impairments in activities of daily living, and deterioration of mental and sexual health. Management depends on the type of incontinence, its aetiology, the severity of symptoms, the effects on QoL, and patient factors. Treatment options include active treatment and passive containment. However, not all active treatment options are feasible in residential care. There is little evidence to advise on standard best practice. OBJECTIVE: The aims of this article are to review treatment options for urinary incontinence in residential care, feasibility of service delivery and challenges associated with this. DISCUSSION: A greater understanding of the issues surrounding the management of urinary incontinence in residential care is required to deliver satisfactory patient-centred care on a consistent basis.

Impact of a novel community-based lifestyle intervention program on type 2 diabetes and cardiovascular risk in a resource-poor setting in the Dominican Republic.

BACKGROUND: The prevalence of type 2 diabetes is increasing at an alarming rate in Latin America and in the Caribbean. We present evidence that a cost-effective and sustainable approach to lifestyle modification in underdeveloped countries can be implemented using community members as healthcare champions. METHODS: Of 222 community members screened in the impoverished community of Villa Juana, Santo Domingo, 69% had prediabetes or diabetes and 79 of these were enrolled. In a 1-day session, trained lifestyle educators from the USA trained ten lay community members to lead groups oriented to lifestyle change. Community leaders met with assigned patient groups monthly for 1 year. Glycated hemoglobin (HbA1c; average plasma glucose concentration), blood pressure, weight and waist circumference were measured at baseline, 6 months and 1 year. RESULTS: 59 individuals completed follow-up. Patients showed significant improvements after 6 months in systolic blood pressure (p=0.001), diastolic blood pressure (p=0.000002) and HbA1c (p=0.015). HbA1c improved further at 1 year (p=0.005). CONCLUSION: Our pilot experience demonstrates the efficacy and feasibility of a novel, low-cost, community-based educational initiative to improve blood glucose control and reduce cardiovascular risk in individuals with type 2 diabetes or prediabetes. Replication of our model in other underserved areas could have a substantial impact on global health.

Deformability considerations in filtration of biological cells.

Biological cells are highly sensitive to variation in local pressure because cellular membranes are not rigid. Unlike microbeads, cells deform under pressure or even lyse. In isolating or enriching cells by mechanical filtration, pressure-induced lysis is exacerbated when high local fluidic velocity is present or when a filter reaches its intended capacity. Microfabrication offers new possibilities to design fluidic environments to reduce cellular stress during the filtration process. We describe the underlying biophysics of cellular stress and general solutions to scale up filtration processes for biological cells.

Microfabricating high-aspect-ratio structures in polyurethane-methacrylate (PUMA) disposable microfluidic devices.

We recently reported a new UV-curable polyurethane-methacrylate (PUMA) resin that has excellent qualities as a disposable microfluidic substrate for clinical diagnostic applications. This article discusses strategies to improve the production yield of PUMA chips that contain dense and high-aspect-ratio features, which presents unique challenges in demolding and bonding steps. These fabrication improvements were deployed to produce a microfiltration device that contained closely spaced and high-aspect-ratio columns, suitable for retaining and concentrating cells or beads from a highly diluted suspension.

A new USP Class VI-compliant substrate for manufacturing disposable microfluidic devices.

As microfluidic systems transition from research tools to disposable clinical-diagnostic devices, new substrate materials are needed to meet both the regulatory requirement as well as the economics of disposable devices. This paper introduces a UV-curable polyurethane-methacrylate (PUMA) substrate that has been qualified for medical use and meets all of the challenges of manufacturing microfluidic devices. PUMA is optically transparent, biocompatible, and exhibits high electroosmotic mobility without surface modification. We report two production processes that are compatible with the existing methods of rapid prototyping and present characterizations of the resultant PUMA microfluidic devices.

Compartmentalization of chemically separated components into droplets.

Not merely a drop in the ocean: The integration of capillary electrophoresis (CE) with droplet generation driven by electroosmotic flow enabled the compartimentalization of molecular components separated by CE in a series of droplets (see picture; the green bars represent the separated analytes). The droplet-confined bands can be docked and studied on a chip.

Simultaneous generation of multiple aqueous droplets in a microfluidic device.

This paper describes a microfluidic platform for the on-demand generation of multiple aqueous droplets, with varying chemical contents or chemical concentrations, for use in droplet based experiments. This generation technique was developed as a complement to existing techniques of continuous-flow (streaming) and discrete-droplet generation by enabling the formation of multiple discrete droplets simultaneously. Here sets of droplets with varying chemical contents can be generated without running the risk of cross-contamination due to the isolated nature of each supply inlet. The use of pressure pulses to generate droplets in parallel is described, and the effect of droplet size is examined in the context of flow rates and surfactant concentrations. To illustrate this technique, an array of different dye-containing droplets was generated, as well as a set of droplets that displayed a concentration gradient of a fluorescent dye.

Fast initiation of chemical reactions with laser-induced breakdown of a nanoscale partition.

This letter describes a new strategy for initiating a chemical reaction that is based on the laser-induced breakdown of a nanoscopic barrier, which physically separates the reactants in space. Because the breakdown of the barrier is fast ( approximately 0.3 micros) and owing to the nanometer dimension of the barrier, the reactants can be brought together and the reaction can be initiated rapidly. The time scale most suited for this method (from microseconds to tens of milliseconds) bridges nicely between the faster time scales that are accessible mostly with laser-based triggering experiments and the slower time scales that are studied most frequently with flow-based devices.

Parametric investigation on the effect of channel topologies on electrophoretic separations.

This paper presents a systematic study that illustrates the importance of the topologies of microchannels on electrokinetically based separation. Using theoretical and numerical analyses, we designed and showed that topologies that significantly increased the surface-to-volume ratio of the channel can provide dramatic improvement in the ability of the channel both to dissipate the heat generated by Joule heating and to reduce the axial dispersion associated with the siphoning effect. The incremental benefit and tradeoff of geometric complexity was also evaluated. The improvement offered by topographically patterned channels, such as finned structures, is especially pertinent in the development of preparative or semi-preparative scale electrokinetically driven separations, such as capillary electrophoresis and capillary electrochromatography, in which large cross sections of channels are required to achieve the needed volumetric throughput.

Microfluidic systems: high radial acceleration in microvortices.

Microfluidic systems can conveniently be used for rapid analysis of biological samples. Here we describe a single re-circulating flow, or microvortex, that can generate a maximum fluid rotational velocity of up to 12 m s(-1) and a corresponding radial acceleration in excess of 10(6)g. Such microvortices may be exploited in centrifugal microdevices to investigate the effects of high radial acceleration on biological and chemical processes.

Fabrication of thermoset polyester microfluidic devices and embossing masters using rapid prototyped polydimethylsiloxane molds.

Plastics are increasingly being used for the fabrication of Lab-on-a-Chip devices due to the variety of beneficial material properties, affordable cost, and straightforward fabrication methods available from a range of different types of plastics. Rapid prototyping of polydimethylsiloxane (PDMS) devices has become a well-known process for the quick and easy fabrication of microfluidic devices in the research laboratory; however, PDMS is not always an appropriate material for every application. This paper describes the fabrication of thermoset polyester microfluidic devices and masters for hot embossing using replica molding techniques. Rapid prototyped PDMS molds are convienently used for the production of non-PDMS polymeric devices. The recessed features in the cast polyester can be bonded to a second polyester piece to form an enclosed microchannel. Thermoset polyester can withstand moderate amounts of pressure and elevated temperature; therefore, the cast polyester piece also can be used as a master for embossing polymethylmethacrylate (PMMA) microfluidic systems. Examples of enclosed polyester and PMMA microchannels are presented, and we discuss the electroosmotic properties of both types of channels, which are important for analytical applications such as capillary electrophoresis.