A5008189820- Wireless Communication Security Techniques
- COVID-19 diagnosis using AI
- Cooperative Communication and Network Coding
- Cellular Automata and Applications
- Radiomics and Machine Learning in Medical Imaging
- Chaos-based Image/Signal Encryption
- Cognitive Radio Networks and Spectrum Sensing
- Lung Cancer Diagnosis and Treatment
- Tuberculosis Research and Epidemiology
- Distributed Sensor Networks and Detection Algorithms
- Error Correcting Code Techniques
- DNA and Biological Computing
- Algorithms and Data Compression
- Artificial Intelligence in Healthcare and Education
- Machine Learning in Healthcare
- Infectious Diseases and Tuberculosis
- Mathematical Analysis and Transform Methods
- Advanced Adaptive Filtering Techniques
- Stability and Control of Uncertain Systems
- Radiology practices and education
- Urinary Tract Infections Management
- Wireless Communication Networks Research
- Urinary Bladder and Prostate Research
- Sparse and Compressive Sensing Techniques
- Smart Systems and Machine Learning
Google (United States)
2012-2025
University of California, Berkeley
2006-2019
BackgroundDeep learning has the potential to augment use of chest radiography in clinical radiology, but challenges include poor generalizability, spectrum bias, and difficulty comparing across studies.PurposeTo develop evaluate deep models for radiograph interpretation by using radiologist-adjudicated reference standards.Materials MethodsDeep were developed detect four findings (pneumothorax, opacity, nodule or mass, fracture) on frontal radiographs. This retrospective study used two data...
Background The World Health Organization (WHO) recommends chest radiography to facilitate tuberculosis (TB) screening. However, radiograph interpretation expertise remains limited in many regions. Purpose To develop a deep learning system (DLS) detect active pulmonary TB on radiographs and compare its performance that of radiologists. Materials Methods A DLS was trained tested using retrospective (acquired between 1996 2020) from 10 countries. improve generalization, large-scale pretraining,...
Background Developing deep learning models for radiology requires large data sets and substantial computational resources. Data set size limitations can be further exacerbated by distribution shifts, such as rapid changes in patient populations standard of care during the COVID-19 pandemic. A common partial mitigation is transfer pretraining a "generic network" on nonmedical then fine-tuning task-specific set. Purpose To reduce requirements chest radiography using an advanced machine...
Chest radiography (CXR) is the most widely-used thoracic clinical imaging modality and crucial for guiding management of cardiothoracic conditions. The detection specific CXR findings has been main focus several artificial intelligence (AI) systems. However, wide range possible abnormalities makes it impractical to build systems detect every condition. In this work, we developed evaluated an AI system classify CXRs as normal or abnormal. For development, used a de-identified dataset 248,445...
We study a problem motivated by cognitive radio in which the primary is packet system that employs ARQ feedback. A secondary allowed to transmit same frequency band provided it ensures attains specified target rate. That is, has certain "interference budget". The crux of does not know how much interference creates on and therefore ignorant its budget. Absent this knowledge, we propose scheme eavesdrops primary's uses knowledge stay within Under assumptions, show there exists an optimal...
Alice and Bob want to share a secret key communicate an independent message, both of which they desire be kept from eavesdropper Eve. We study this problem communication generation when two resources are available -- correlated sources at Alice, Bob, Eve, noisy broadcast channel Eve is the sources. interested in characterizing fundamental trade-off between rates message key. present achievable solution prove its optimality for parallel channels case each sub-channel source component...
The utility of limited feedback for coding over an individual sequence discrete memoryless channels is investigated. This study complements recent results showing how or noisy can boost the reliability communication. A strategy with fixed input distribution <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">P</i> given that asymptotically achieves rates arbitrarily close to mutual information induced by and state-averaged channel. When...
In this work, we present an approach, which call Embeddings for Language/Image-aligned X-Rays, or ELIXR, that leverages a language-aligned image encoder combined grafted onto fixed LLM, PaLM 2, to perform broad range of chest X-ray tasks. We train lightweight adapter architecture using images paired with corresponding free-text radiology reports from the MIMIC-CXR dataset. ELIXR achieved state-of-the-art performance on zero-shot (CXR) classification (mean AUC 0.850 across 13 findings),...
Alice and Bob want to share a secret key communicate an independent message, both of which they desire be kept from eavesdropper Eve. We study this problem communication generation when two resources are available — correlated sources at Alice, Bob, Eve, noisy broadcast channel No other resource, in particular, no is available. interested characterizing the fundamental trade-off between rates message key. present achievable solution based on separation architecture prove its optimality under...
"Just Accepted" papers have undergone full peer review and been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, proof before it is published its final version. Please note that during production of the copyedited article, errors may be discovered which could affect content. Purpose To evaluate impact an artificial intelligence (AI) assistant lung cancer screening (LCS) on multinational clinical workflows. Materials Methods An AI...
A fundamental problem in dynamic frequency reuse is that the cognitive radio ignorant of amount interference it inflicts on primary license holder. Policies attempt to limit without active participation are thus difficult implement. However, many wireless systems use flow control feedback such as ARQs. By listening these signals, a can obtain indirect information about generates and behave an acceptable manner. This paper introduces information-theoretic model this basic observation develops...
The distributed remote source coding (the so-called CEO) problem is studied in the case where underlying source, not necessarily Gaussian, has finite differential entropy and observation noise Gaussian. main result a new lower bound for sum-rate-distortion function under arbitrary distortion measures. When specialized to of mean-squared error, it shown that exactly mirrors corresponding upper bound, except power (variance), whereas power. Bounds exhibiting this pleasing duality have been...
A generalization of the additive Gaussian two-way channel to <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">M</i> users is considered. Such channels contain implicit feedback in sense that output signals observed by different encoders are correlated. While benefits shown be negligible at high SNR, for moderate can play a significant role boosting sum-rate performance. To highlight this potential gain, special case -user multiway with common By...
Recently, Shayevitz and Feeler introduced an individual sequence formulation of channel coding under model uncertainty elegant strategy that adapts Horstein's scheme to this setting achieve the empirical capacity channel. Their requires both full-rate output feedback common randomness. We present a in style Hybrid ARQ no by using randomness zero-rate active feedback. This still asymptotically achieves capacity.
Using artificial intelligence (AI) to interpret chest X-rays (CXRs) could support accessible triage tests for active pulmonary tuberculosis (TB) in resource-constrained settings. The performance of two cloud-based CXR AI systems - one detect TB and the other abnormalities a population with high human immunodeficiency virus (HIV) burden was evaluated. We recruited 1978 adults who had symptoms, were close contacts known patients, or newly diagnosed HIV at three clinical sites. TB-detecting (TB...
In the CEO problem, introduced by Berger et al, IEEE Trans. Info. Theory, 1996, a is interested in source that cannot be observed directly. M agents observe independently noisy versions of and, without collaborating, must encode these across noiseless rate-constrained channels to CEO. The quadratic AWGN problem refers class problems for which view through additive white Gaussian noise, and distortion squared error. This paper discusses two upper bounds sum-rate function this problems. first...
Encoding correlated sources at separate encoders has been studied extensively from the perspective of asymptotically long block codes. The associated error exponents are known for case lossless source coding. In this paper, we introduce a novel technique deriving achievable lossy coding problems, where original need to be reconstructed within some fidelity. As an example, show how apply our determine Berger-Yeung problem.
Alice wants to send Bob a secret message in the presence of an eavesdropper Eve. has two resources available: noisy broadcast channel and source observations correlated with at We consider cases, first which are independent second they can depend on each other. For case, we strategy that exploits sources channels separately show this separation is optimal for class includes Gaussian example. case other, does not work, yet be established. setting, improve best known upper bound capacity.