-
Self-Supervised Deep Learning to Enhance Breast Cancer Detection on Screening Mammography
Authors:
John D. Miller,
Vignesh A. Arasu,
Albert X. Pu,
Laurie R. Margolies,
Weiva Sieh,
Li Shen
Abstract:
A major limitation in applying deep learning to artificial intelligence (AI) systems is the scarcity of high-quality curated datasets. We investigate strong augmentation based self-supervised learning (SSL) techniques to address this problem. Using breast cancer detection as an example, we first identify a mammogram-specific transformation paradigm and then systematically compare four recent SSL m…
▽ More
A major limitation in applying deep learning to artificial intelligence (AI) systems is the scarcity of high-quality curated datasets. We investigate strong augmentation based self-supervised learning (SSL) techniques to address this problem. Using breast cancer detection as an example, we first identify a mammogram-specific transformation paradigm and then systematically compare four recent SSL methods representing a diversity of approaches. We develop a method to convert a pretrained model from making predictions on uniformly tiled patches to whole images, and an attention-based pooling method that improves the classification performance. We found that the best SSL model substantially outperformed the baseline supervised model. The best SSL model also improved the data efficiency of sample labeling by nearly 4-fold and was highly transferrable from one dataset to another. SSL represents a major breakthrough in computer vision and may help the AI for medical imaging field to shift away from supervised learning and dependency on scarce labels.
△ Less
Submitted 15 March, 2022;
originally announced March 2022.
-
Deep Learning to Improve Breast Cancer Early Detection on Screening Mammography
Authors:
Li Shen,
Laurie R. Margolies,
Joseph H. Rothstein,
Eugene Fluder,
Russell B. McBride,
Weiva Sieh
Abstract:
The rapid development of deep learning, a family of machine learning techniques, has spurred much interest in its application to medical imaging problems. Here, we develop a deep learning algorithm that can accurately detect breast cancer on screening mammograms using an "end-to-end" training approach that efficiently leverages training datasets with either complete clinical annotation or only the…
▽ More
The rapid development of deep learning, a family of machine learning techniques, has spurred much interest in its application to medical imaging problems. Here, we develop a deep learning algorithm that can accurately detect breast cancer on screening mammograms using an "end-to-end" training approach that efficiently leverages training datasets with either complete clinical annotation or only the cancer status (label) of the whole image. In this approach, lesion annotations are required only in the initial training stage, and subsequent stages require only image-level labels, eliminating the reliance on rarely available lesion annotations. Our all convolutional network method for classifying screening mammograms attained excellent performance in comparison with previous methods. On an independent test set of digitized film mammograms from Digital Database for Screening Mammography (DDSM), the best single model achieved a per-image AUC of 0.88, and four-model averaging improved the AUC to 0.91 (sensitivity: 86.1%, specificity: 80.1%). On a validation set of full-field digital mammography (FFDM) images from the INbreast database, the best single model achieved a per-image AUC of 0.95, and four-model averaging improved the AUC to 0.98 (sensitivity: 86.7%, specificity: 96.1%). We also demonstrate that a whole image classifier trained using our end-to-end approach on the DDSM digitized film mammograms can be transferred to INbreast FFDM images using only a subset of the INbreast data for fine-tuning and without further reliance on the availability of lesion annotations. These findings show that automatic deep learning methods can be readily trained to attain high accuracy on heterogeneous mammography platforms, and hold tremendous promise for improving clinical tools to reduce false positive and false negative screening mammography results.
△ Less
Submitted 31 December, 2018; v1 submitted 30 August, 2017;
originally announced August 2017.
-
Provisioning Low Latency, Resilient Mobile Edge Clouds for 5G
Authors:
Russell Ford,
Ashwin Sridharan,
Robert Margolies,
Rittwik Jana,
Sundeep Rangan
Abstract:
Network virtualization and SDN-based routing allow carriers to flexibly configure their networks in response to demand and unexpected network disruptions. However, cellular networks, by nature, pose some unique challenges because of user mobility and control/data plane partitioning, which calls for new architectures and provisioning paradigms. In this paper, we address the latter part by devising…
▽ More
Network virtualization and SDN-based routing allow carriers to flexibly configure their networks in response to demand and unexpected network disruptions. However, cellular networks, by nature, pose some unique challenges because of user mobility and control/data plane partitioning, which calls for new architectures and provisioning paradigms. In this paper, we address the latter part by devising algorithms that can provision the data plane to create a distributed Mobile Edge Cloud (MEC), which provides opportunities for lower latencies and increased resilience (through placement of network functions at more distributed datacenter locations) and accounts for service disruption that could be incurred because of user mobility between the service areas of different datacenters. Through evaluations with topology and traffic data from a major carriers's network, we show that, compared to static, centralized networks, careful virtualized provisioning can yield significant savings in network costs while still minimizing service disruption due to mobility. We demonstrate that up to a 75% reduction in redundant datacenter capacity over the operator's current topology (while achieving the same level of resilience) is possible by distributing load over many mobile cloud datacenters.
△ Less
Submitted 31 March, 2017;
originally announced March 2017.
-
WiLiTV: A Low-Cost Wireless Framework for Live TV Services
Authors:
Rajeev Kumar,
Robert S Margolies,
Rittwik Jana,
Yong Liu,
Shivendra Panwar
Abstract:
With the evolution of HDTV and Ultra HDTV, the bandwidth requirement for IP-based TV content is rapidly increasing. Consumers demand uninterrupted service with a high Quality of Experience (QoE). Service providers are constantly trying to differentiate themselves by innovating new ways of distributing content more efficiently with lower cost and higher penetration. In this work, we propose a cost-…
▽ More
With the evolution of HDTV and Ultra HDTV, the bandwidth requirement for IP-based TV content is rapidly increasing. Consumers demand uninterrupted service with a high Quality of Experience (QoE). Service providers are constantly trying to differentiate themselves by innovating new ways of distributing content more efficiently with lower cost and higher penetration. In this work, we propose a cost-efficient wireless framework (WiLiTV) for delivering live TV services, consisting of a mix of wireless access technologies (e.g. Satellite, WiFi and LTE overlay links). In the proposed architecture, live TV content is injected into the network at a few residential locations using satellite dishes. The content is then further distributed to other homes using a house-to-house WiFi network or via an overlay LTE network. Our problem is to construct an optimal TV distribution network with the minimum number of satellite injection points, while preserving the highest QoE, for different neighborhood densities. We evaluate the framework using realistic time-varying demand patterns and a diverse set of home location data. Our study demonstrates that the architecture requires 75 - 90% fewer satellite injection points, compared to traditional architectures. Furthermore, we show that most cost savings can be obtained using simple and practical relay routing solutions.
△ Less
Submitted 10 January, 2017;
originally announced January 2017.
-
Panda: Neighbor Discovery on a Power Harvesting Budget
Authors:
Robert Margolies,
Guy Grebla,
Tingjun Chen,
Dan Rubenstein,
Gil Zussman
Abstract:
Object tracking applications are gaining popularity and will soon utilize Energy Harvesting (EH) low-power nodes that will consume power mostly for Neighbor Discovery (ND) (i.e., identifying nodes within communication range). Although ND protocols were developed for sensor networks, the challenges posed by emerging EH low-power transceivers were not addressed. Therefore, we design an ND protocol t…
▽ More
Object tracking applications are gaining popularity and will soon utilize Energy Harvesting (EH) low-power nodes that will consume power mostly for Neighbor Discovery (ND) (i.e., identifying nodes within communication range). Although ND protocols were developed for sensor networks, the challenges posed by emerging EH low-power transceivers were not addressed. Therefore, we design an ND protocol tailored for the characteristics of a representative EH prototype: the TI eZ430-RF2500-SEH. We present a generalized model of ND accounting for unique prototype characteristics (i.e., energy costs for transmission/reception, and transceiver state switching times/costs). Then, we present the Power Aware Neighbor Discovery Asynchronously (Panda) protocol in which nodes transition between the sleep, receive, and transmit states. We analyze \name and select its parameters to maximize the ND rate subject to a homogeneous power budget. We also present Panda-D, designed for non-homogeneous EH nodes. We perform extensive testbed evaluations using the prototypes and study various design tradeoffs. We demonstrate a small difference (less then 2%) between experimental and analytical results, thereby confirming the modeling assumptions. Moreover, we show that Panda improves the ND rate by up to 3x compared to related protocols. Finally, we show that Panda-D operates well under non-homogeneous power harvesting.
△ Less
Submitted 24 January, 2016;
originally announced January 2016.
-
Project-based Learning within a Large-Scale Interdisciplinary Research Effort
Authors:
Robert Margolies,
Maria Gorlatova,
John Sarik,
Peter Kinget,
Ioannis Kymissis,
Gil Zussman
Abstract:
The modern engineering landscape increasingly requires a range of skills to successfully integrate complex systems. Project-based learning is used to help students build professional skills. However, it is typically applied to small teams and small efforts. This paper describes an experience in engaging a large number of students in research projects within a multi-year interdisciplinary research…
▽ More
The modern engineering landscape increasingly requires a range of skills to successfully integrate complex systems. Project-based learning is used to help students build professional skills. However, it is typically applied to small teams and small efforts. This paper describes an experience in engaging a large number of students in research projects within a multi-year interdisciplinary research effort. The projects expose the students to various disciplines in Computer Science (embedded systems, algorithm design, networking), Electrical Engineering (circuit design, wireless communications, hardware prototy**), and Applied Physics (thin-film battery design, solar cell fabrication). While a student project is usually focused on one discipline area, it requires interaction with at least two other areas. Over 5 years, 180 semester-long projects have been completed. The students were a diverse group of high school, undergraduate, and M.S. Computer Science, Computer Engineering, and Electrical Engineering students. Some of the approaches that were taken to facilitate student learning are real-world system development constraints, regular cross-group meetings, and extensive involvement of Ph.D. students in student mentorship and knowledge transfer. To assess the approaches, a survey was conducted among the participating students. The results demonstrate the effectiveness of the approaches. For example, 70% of the students surveyed indicated that working on their research project improved their ability to function on multidisciplinary teams more than coursework, internships, or any other activity.
△ Less
Submitted 25 October, 2014;
originally announced October 2014.
