Search

A comprehensive overview of medical AI devices approved by the US Food and Drug Administration sheds new light on limitations of the evaluation process that can mask vulnerabilities of devices when they are deployed on patients.

Breast cancer remains a global challenge, causing over 600,000 deaths in 2018 (ref. 1). To achieve earlier cancer detection, health organizations worldwide recommend screening mammography, which is estimated to decrease breast cancer mortality by 20–40% (refs. 2,3). Despite the clear value of screening mammography, significant false positive and false negative rates along with non-uniformities in expert reader availability leave opportunities for improving quality and access4,5. To address these limitations, there has been much recent interest in applying deep learning to mammography6–18, and these efforts have highlighted two key difficulties: obtaining large amounts of annotated training data and ensuring generalization across populations, acquisition equipment and modalities. Here we present an annotation-efficient deep learning approach that (1) achieves state-of-the-art performance in mammogram classification, (2) successfully extends to digital breast tomosynthesis (DBT; ‘3D mammography’), (3) detects cancers in clinically negative prior mammograms of patients with cancer, (4) generalizes well to a population with low screening rates and (5) outperforms five out of five full-time breast-imaging specialists with an average increase in sensitivity of 14%. By creating new ‘maximum suspicion projection’ (MSP) images from DBT data, our progressively trained, multiple-instance learning approach effectively trains on DBT exams using only breast-level labels while maintaining localization-based interpretability. Altogether, our results demonstrate promise towards software that can improve the accuracy of and access to screening mammography worldwide. A generalizable and interpretable artificial-intelligence system achieves clinical accuracy for screening and early breast-cancer detection on 2D and 3D mammograms.

Computerized automatic methods have been employed to boost the productivity as well as objectiveness of hand bone age assessment. These approaches make predictions according to the whole X-ray images, which include other objects that may introduce distractions. Instead, our framework is inspired by the clinical workflow (Tanner-Whitehouse) of hand bone age assessment, which focuses on the key components of the hand. The proposed framework is composed of two components: a Mask R-CNN subnet of pixelwise hand segmentation and a residual attention network for hand bone age assessment. The Mask R-CNN subnet segments the hands from X-ray images to avoid the distractions of other objects (e.g., X-ray tags). The hierarchical attention components of the residual attention subnet force our network to focus on the key components of the X-ray images and generate the final predictions as well as the associated visual supports, which is similar to the assessment procedure of clinicians. We evaluate the performance of the proposed pipeline on the RSNA pediatric bone age dataset 1 and the results demonstrate its superiority over the previous methods.1http://rsnachallenges.cloudapp.net/competitions/4

Advancements in convolutional neural networks (CNNs) have made significant strides toward achieving high performance levels on multiple object recognition tasks. While some approaches utilize information from the entire scene to propose regions of interest, the task of interpreting a particular region or object is still performed independently of other objects and features in the image. Here we demonstrate that a scene's ‘gist’ can significantly contribute to how well humans can recognize objects. These findings are consistent with the notion that humans foveate on an object and incorporate information from the periphery to aid in recognition. We use a biologically inspired two-part convolutional neural network ('GistNet') that models the fovea and periphery to provide a proof-of-principle demonstration that computational object recognition can significantly benefit from the gist of the scene as contextual information. Our model yields accuracy improvements of up to 50% in certain object categories when incorporating contextual gist, while only increasing the original model size by 5%. This proposed model mirrors our intuition about how the human visual system recognizes objects, suggesting specific biologically plausible constraints to improve machine vision and building initial steps towards the challenge of scene understanding.

Deep learning approaches to breast cancer detection in mammograms have recently shown promising results. However, such models are constrained by the limited size of publicly available mammography datasets, in large part due to privacy concerns and the high cost of generating expert annotations. Limited dataset size is further exacerbated by substantial class imbalance since “normal” images dramatically outnumber those with findings. Given the rapid progress of generative models in synthesizing realistic images, and the known effectiveness of simple data augmentation techniques (e.g. horizontal flipping), we ask if it is possible to synthetically augment mammogram datasets using generative adversarial networks (GANs). We train a class-conditional GAN to perform contextual in-filling, which we then use to synthesize lesions onto healthy screening mammograms. First, we show that GANs are capable of generating high-resolution synthetic mammogram patches. Next, we experimentally evaluate using the augmented dataset to improve breast cancer classification performance. We observe that a ResNet-50 classifier trained with GAN-augmented training data produces a higher AUROC compared to the same model trained only on traditionally augmented data, demonstrating the potential of our approach.

Laser scanning cytometry (LSC) is emerging as a clinical tool. In one application a "Clatch" slide, named after the inventor, is used in conjunction with LSC for cell surface marker immunophenotyping of patient samples. The slide requires time consuming and laborious pipetting steps, making a test tedious and prone to handling errors. The Clatch slide also uses a significant number of cells, limiting the number of analyses on paucicellular samples. This paper presents an automated microfluidic system consisting of a control circuit, a microfluidic system, and an aluminum frame, capable of performing immunophenotyping procedures. This prototype system reduces 36 pipetting steps to 1, reduces the amount of cell sample from 180 μL to 56 μL, and shortens the time used by technicians.

The REQS package is an interface between the R environment of statistical computing and the EQS software for structural equation modeling. The package consists of 3 main functions that read EQS script files and import the results into R, call EQS script files from R, and run EQS script files from R and import the results after EQS computations. The components a user needs are R and EQS. We give a short introduction to R and EQS, elaborate the functionalities of the package, and show how to use the package by means of several examples with special emphasis on simulations. The first simulation investigates the effects of nonnormal data on various test statistics. The second simulation conducts a power analysis for a path coefficient in a 2-group model.

As the industry drives to lower k1 imaging we commonly accept the use of higher NA imaging and advanced illumination conditions. The advent of this technology shift has given rise to very exotic pupil spread functions that have some areas of high thermal energy density creating new modeling and control challenges. Modern scanners are equipped with advanced lens manipulators that introduce controlled adjustments of the lens elements to counteract the lens aberrations existing in the system. However, there are some specific non-correctable aberration modes that are detrimental to important structures. In this paper, we introduce a methodology for minimizing the impact of aberrations for specific designs at hand. We employ computational lithography to analyze the design being imaged, and then devise a lens manipulator control scheme aimed at optimizing the aberration level for the specific design. The optimization scheme does not minimize the overall aberration, but directs the aberration control to optimize the imaging performance, such as CD control or process window, for the target design. Through computational lithography, we can identify the aberration modes that are most detrimental to the design, and also correlations between imaging responses of independent aberration modes. Then an optimization algorithm is applied to determine how to use the lens manipulators to drive the aberrations modes to levels that are best for the specified imaging performance metric achievable with the tool. We show an example where this method is applied to an aggressive memory device imaged with an advanced ArF scanner. We demonstrate with both simulation and experimental data that this application specific tool optimization successfully compensated for the thermal induced aberrations dynamically, improving the imaging performance consistently through the lot.

Accounting is critical for information technology budgeting and chargeback. Traditional accounting in UNIX/Linux systems is known as process accounting, in which an accounting record is created when a process ends. System administrators typically aggregate accounting records based on individual users or groups. As Web and application servers along with databases handle requests and transactions for multiple entities in various Web applications and services, LPAR accounting and transaction accounting become increasingly critical for service providers in shared resource environments. In this paper we present the design and implementation of a J2EE accounting application for resource usage metering. For process accounting the resulting system can generate usage reports by projects, by groups, by users, by commands, or by a combination of these identifiers. For dynamically changing partitions it generates reports for shared resources including CPUs, memories, disks, file systems, and network interfaces. For transaction accounting it generates reports based on account classes provided that applications are instrumented. It is the first known J2EE accounting application for UNIX/Linux transaction accounting.

Learning on the Internet becomes the main stream in the deployment of distance education and training. In such an environment, both learners and teachers or peer-to-peer are able to join the learning activities synchronously as well as asynchronously. However, the present user interfaces still do not provide enough flexibility and interactively to both learners and teachers. As a result, the users may lack of interests to continuously use the prepared learning environment. In this paper, we design and implement a multimedia user interface, which contains video, audio, text contents, and authoring tools. We integrate the state-of-the-art technologies, such as image segmentation, object tracking, and voice over IP. In the video portion, the system is able to automatically track the teacher movement. Teachers have high degree of freedom in presenting themselves to attract learners' attention. To achieve the above, the developed region-based image segmentation scheme and tracking scheme is realized in real-time. Special voice transmission scheme based on forward error correction is also designed to wrestle with the transmission difficulties on the present Internet. Authoring tools are provided to teachers and learners to improve the convenient utilization of the system.

To develop optimal parallel I/O subsystems, one must have a thorough understanding of the workload characteristics of parallel I/O and its exploitation of the associated parallel file system. Presented are the results of a study conducted to analyze the parallel I/O workloads of several applications on a parallel processor using the Vesta parallel file system. Traces of the applications are obtained to collect system events, communication events, and parallel I/O events. The traces are then analyzed to determine workload characteristics. The results show I/O request rates on the order of hundreds of requests per second, a large majority of requests are for small amounts of data (less than 1500 bytes), a few requests are for large amounts of data (on the order of megabytes), significant file sharing among processes within a job, and strong temporal, traditional spatial, and interprocess spatial locality.

We propose efficient stack simulation algorithms for shared memory multiprocessor (MP) caches. A stack simulation algorithm for write-updated MP caches is first presented. It produces the number of write-updates as well as misses for all cache configurations in a single run. We then devise a new stack simulation algorithm for writeinvalidate MP caches. Our algorithm takes into account cross-invalidation among processors, and generates the number of invalidations as well as misses for all cache configurations in a single run. A cache simulator based on our algorithms for MP caches is developed and the results on sample traces are reported. Our results show that effi cient stack simulation is a powerful technique for multi processor cache analysis.