Browsing by Subject "Artificial Intelligence"
Now showing 1 - 10 of 32
Results Per Page
Sort Options
Item AI on the Edge with CondenseNeXt: An Efficient Deep Neural Network for Devices with Constrained Computational Resources(2021-08) Kalgaonkar, Priyank B.; El-Sharkawy, Mohamed A.; King, Brian S.; Rizkalla, Maher E.Research work presented within this thesis propose a neoteric variant of deep convolutional neural network architecture, CondenseNeXt, designed specifically for ARM-based embedded computing platforms with constrained computational resources. CondenseNeXt is an improved version of CondenseNet, the baseline architecture whose roots can be traced back to ResNet. CondeseNeXt replaces group convolutions in CondenseNet with depthwise separable convolutions and introduces group-wise pruning, a model compression technique, to prune (remove) redundant and insignificant elements that either are irrelevant or do not affect performance of the network upon disposition. Cardinality, a new dimension to the existing spatial dimensions, and class-balanced focal loss function, a weighting factor inversely proportional to the number of samples, has been incorporated in order to relieve the harsh effects of pruning, into the design of CondenseNeXt’s algorithm. Furthermore, extensive analyses of this novel CNN architecture was performed on three benchmarking image datasets: CIFAR-10, CIFAR-100 and ImageNet by deploying the trained weight on to an ARM-based embedded computing platform: NXP BlueBox 2.0, for real-time image classification. The outputs are observed in real-time in RTMaps Remote Studio’s console to verify the correctness of classes being predicted. CondenseNeXt achieves state-of-the-art image classification performance on three benchmark datasets including CIFAR-10 (4.79% top-1 error), CIFAR-100 (21.98% top-1 error) and ImageNet (7.91% single model, single crop top-5 error), and up to 59.98% reduction in forward FLOPs compared to CondenseNet. CondenseNeXt can also achieve a final trained model size of 2.9 MB, however at the cost of 2.26% in accuracy loss. Thus, performing image classification on ARM-Based computing platforms without requiring a CUDA enabled GPU support, with outstanding efficiency.Item Analysis of Latent Space Representations for Object Detection(2024-08) Dale, Ashley Susan; Christopher, Lauren; King, Brian; Salama, Paul; Rizkalla, MaherDeep Neural Networks (DNNs) successfully perform object detection tasks, and the Con- volutional Neural Network (CNN) backbone is a commonly used feature extractor before secondary tasks such as detection, classification, or segmentation. In a DNN model, the relationship between the features learned by the model from the training data and the features leveraged by the model during test and deployment has motivated the area of feature interpretability studies. The work presented here applies equally to white-box and black-box models and to any DNN architecture. The metrics developed do not require any information beyond the feature vector generated by the feature extraction backbone. These methods are therefore the first methods capable of estimating black-box model robustness in terms of latent space complexity and the first methods capable of examining feature representations in the latent space of black box models. This work contributes the following four novel methodologies and results. First, a method for quantifying the invariance and/or equivariance of a model using the training data shows that the representation of a feature in the model impacts model performance. Second, a method for quantifying an observed domain gap in a dataset using the latent feature vectors of an object detection model is paired with pixel-level augmentation techniques to close the gap between real and synthetic data. This results in an improvement in the model’s F1 score on a test set of outliers from 0.5 to 0.9. Third, a method for visualizing and quantifying similarities of the latent manifolds of two black-box models is used to correlate similar feature representation with increase success in the transferability of gradient-based attacks. Finally, a method for examining the global complexity of decision boundaries in black-box models is presented, where more complex decision boundaries are shown to correlate with increased model robustness to gradient-based and random attacks.Item Artificial Intelligence Approaches to Assessing Primary Cilia(MyJove Corp., 2021-05-01) Bansal, Ruchi; Engle, Staci E.; Kamba, Tisianna K.; Brewer, Kathryn M.; Lewis, Wesley R.; Berbari, Nicolas F.; Biology, School of ScienceCilia are microtubule based cellular appendages that function as signaling centers for a diversity of signaling pathways in many mammalian cell types. Cilia length is highly conserved, tightly regulated, and varies between different cell types and tissues and has been implicated in directly impacting their signaling capacity. For example, cilia have been shown to alter their lengths in response to activation of ciliary G protein-coupled receptors. However, accurately and reproducibly measuring the lengths of numerous cilia is a time-consuming and labor-intensive procedure. Current approaches are also error and bias prone. Artificial intelligence (Ai) programs can be utilized to overcome many of these challenges due to capabilities that permit assimilation, manipulation, and optimization of extensive data sets. Here, we demonstrate that an Ai module can be trained to recognize cilia in images from both in vivo and in vitro samples. After using the trained Ai to identify cilia, we are able to design and rapidly utilize applications that analyze hundreds of cilia in a single sample for length, fluorescence intensity and co-localization. This unbiased approach increased our confidence and rigor when comparing samples from different primary neuronal preps in vitro as well as across different brain regions within an animal and between animals. Moreover, this technique can be used to reliably analyze cilia dynamics from any cell type and tissue in a high-throughput manner across multiple samples and treatment groups. Ultimately, Ai-based approaches will likely become standard as most fields move toward less biased and more reproducible approaches for image acquisition and analysis.Item Artificial Intelligence for Contrast-Free MRI: Scar Assessment in Myocardial Infarction Using Deep Learning-Based Virtual Native Enhancement(American Heart Association, 2022-11-15) Zhang, Qiang; Burrage, Matthew K.; Shanmuganathan, Mayooran; Gonzales, Ricardo A.; Lukaschuk, Elena; Thomas, Katharine E.; Mills, Rebecca; Pelado, Joana Leal; Nikolaidou, Chrysovalantou; Popescu, Iulia A.; Lee, Yung P.; Zhang, Xinheng; Dharmakumar, Rohan; Myerson, Saul G.; Rider, Oliver; Oxford Acute Myocardial Infarction (OxAMI) Study; Channon, Keith M.; Neubauer, Stefan; Piechnik, Stefan K.; Ferreira, Vanessa M.; Medicine, School of MedicineBackground: Myocardial scar is currently assessed non-invasively using cardiovascular magnetic resonance (CMR) late gadolinium enhancement (LGE) as an imaging gold-standard. However, a contrast-free approach would provide many advantages, including a faster and cheaper scan, without contrast-associated problems. Methods: Virtual Native Enhancement (VNE) is a novel technology that can produce virtual LGE-like images, without the need for contrast. VNE combines cine imaging and native T1-maps to produce LGE-like images using artificial intelligence (AI). VNE was developed for patients with prior myocardial infarction on 4271 datasets (912 patients), where each dataset is comprised of slice position-matched cine, T1-maps and LGE images. After quality control, 3002 datasets (775 patients) were used for development, and 291 datasets (68 patients) for testing. The VNE generator was trained using generative adversarial networks, employing two adversarial discriminators to improve the image quality. The left ventricle was contoured semi-automatically. Myocardial scar volume was quantified using the full width at half maximum method. Scar transmurality was measured using the centerline chord method and visualized on bull’s eye plots. Lesion quantification by VNE and LGE were compared using linear regression, Pearson correlation (R) and intraclass correlation coefficients (ICC). Proof-of-principle histopathological comparison of VNE in a porcine model of myocardial infarction was also performed. Results: VNE provided significantly better image quality than LGE on blinded analysis by 5 independent operators on 291 datasets (all p<0.001). VNE correlated strongly with LGE in quantifying scar size (R=0.89, ICC=0.94) and transmurality (R=0.84, ICC=0.90) in 66 patients (277 test datasets). Two CMR experts reviewed all test image slices and reported an overall accuracy of 84% of VNE in detecting scar when compared with LGE, with specificity of 100% and sensitivity of 77%. VNE also showed excellent visuospatial agreement with histopathology in 2 cases of a porcine model of myocardial infarction. Conclusions: VNE demonstrated high agreement with LGE-CMR for myocardial scar assessment in patients with prior myocardial infarction in visuospatial distribution and lesion quantification, with superior image quality. VNE is a potentially transformative AI-based technology, with promise to reduce scan times and costs, increase clinical throughput, and improve the accessibility of CMR in the very-near future.Item Bridging The Gap Between Healthcare Providers and Consumers: Extracting Features from Online Health Forum to Meet Social Needs of Patients using Network Analysis and Embedding(2020-08) Mokashi, Maitreyi; Chakraborty, Sunandan; Jones, Josette; Zheng, JiapingChronic disease patients have to face many issues during and after their treatment. A lot of these issues are either personal, professional, or social in nature. It may so happen that these issues are overlooked by the respective healthcare providers and become major obstacles in the patient’s day-to-day life and their disease management. We extract data from an online health platform that serves as a ‘safe haven’ to the patients and survivors to discuss help and coping issues. This thesis presents a novel approach that acts as the first step to include the social issues discussed by patients on online health forums which the healthcare providers need to consider in order to create holistic treatment plans. There are numerous online forums where patients share their experiences and post questions about their treatments and their subsequent side effects. We collected data from an “Online Breast Cancer Forum”. On this forum, users (patients) have created threads across many related topics and shared their experiences and questions. We connect the patients (users) with the topic in which they have posted by converting the data into a bipartite network and turn the network nodes into a high-dimensional feature space. From this feature space, we perform community detection on the node embeddings to unearth latent connections between patients and topics. We claim that these latent connections, along with the existing ones, will help to create a new knowledge base that will eventually help the healthcare providers to understand and acknowledge the non-medical related issues to a treatment, and create more adaptive and personalized plans. We performed both qualitative and quantitative analysis on the obtained embeddings to prove the superior quality of our approach and its potential to extract more information when compared to other models.Item Building the case for actionable ethics in digital health research supported by artificial intelligence(Springer Nature, 2019-07-17) Nebeker, Camille; Torous, John; Bartlett Ellis, Rebecca J.; School of NursingThe digital revolution is disrupting the ways in which health research is conducted, and subsequently, changing healthcare. Direct-to-consumer wellness products and mobile apps, pervasive sensor technologies and access to social network data offer exciting opportunities for researchers to passively observe and/or track patients ‘in the wild’ and 24/7. The volume of granular personal health data gathered using these technologies is unprecedented, and is increasingly leveraged to inform personalized health promotion and disease treatment interventions. The use of artificial intelligence in the health sector is also increasing. Although rich with potential, the digital health ecosystem presents new ethical challenges for those making decisions about the selection, testing, implementation and evaluation of technologies for use in healthcare. As the ‘Wild West’ of digital health research unfolds, it is important to recognize who is involved, and identify how each party can and should take responsibility to advance the ethical practices of this work. While not a comprehensive review, we describe the landscape, identify gaps to be addressed, and offer recommendations as to how stakeholders can and should take responsibility to advance socially responsible digital health research.Item ChatGPT-3.5 System Usability Scale early assessment among Healthcare Workers: Horizons of adoption in medical practice(Elsevier, 2024-04-07) Aljamaan, Fadi; Malki, Khalid H.; Alhasan, Khalid; Jamal, Amr; Altamimi, Ibraheem; Khayat, Afnan; Alhaboob, Ali; Abdulmajeed, Naif; Alshahrani, Fatimah S.; Saad, Khaled; Al-Eyadhy, Ayman; Al-Tawfiq, Jaffar A.; Temsah, Mohamad-Hani; Medicine, School of MedicineArtificial intelligence (AI) chatbots, such as ChatGPT, have widely invaded all domains of human life. They have the potential to transform healthcare future. However, their effective implementation hinges on healthcare workers' (HCWs) adoption and perceptions. This study aimed to evaluate HCWs usability of ChatGPT three months post-launch in Saudi Arabia using the System Usability Scale (SUS). A total of 194 HCWs participated in the survey. Forty-seven percent were satisfied with their usage, 57 % expressed moderate to high trust in its ability to generate medical decisions. 58 % expected ChatGPT would improve patients' outcomes, even though 84 % were optimistic of its potential to improve the future of healthcare practice. They expressed possible concerns like recommending harmful medical decisions and medicolegal implications. The overall mean SUS score was 64.52, equivalent to 50 % percentile rank, indicating high marginal acceptability of the system. The strongest positive predictors of high SUS scores were participants' belief in AI chatbot's benefits in medical research, self-rated familiarity with ChatGPT and self-rated computer skills proficiency. Participants' learnability and ease of use score correlated positively but weakly. On the other hand, medical students and interns had significantly high learnability scores compared to others, while ease of use scores correlated very strongly with participants' perception of positive impact of ChatGPT on the future of healthcare practice. Our findings highlight the HCWs' perceived marginal acceptance of ChatGPT at the current stage and their optimism of its potential in supporting them in future practice, especially in the research domain, in addition to humble ambition of its potential to improve patients' outcomes particularly in regard of medical decisions. On the other end, it underscores the need for ongoing efforts to build trust and address ethical and legal concerns of AI implications in healthcare. The study contributes to the growing body of literature on AI chatbots in healthcare, especially addressing its future improvement strategies and provides insights for policymakers and healthcare providers about the potential benefits and challenges of implementing them in their practice.Item Community Recommendation in Social Networks with Sparse Data(2020-12) Rahmaniazad, Emad; King, Brian; Jafari, Ali; Salama, PaulRecommender systems are widely used in many domains. In this work, the importance of a recommender system in an online learning platform is discussed. After explaining the concept of adding an intelligent agent to online education systems, some features of the Course Networking (CN) website are demonstrated. Finally, the relation between CN, the intelligent agent (Rumi), and the recommender system is presented. Along with the argument of three different approaches for building a community recommendation system. The result shows that the Neighboring Collaborative Filtering (NCF) outperforms both the transfer learning method and the Continuous bag-of-words approach. The NCF algorithm has a general format with two various implementations that can be used for other recommendations, such as course, skill, major, and book recommendations.Item A Computer Wrote this Paper: What ChatGPT Means for Education, Research, and Writing(2023-01-26) Bishop, LeaOf particular interest to educators, an exploration of what new language-generation software does (and does not) do well. Argues that the new language-generation models make instruction in writing mechanics irrelevant, and that educators should shift to teaching only the more advanced writing skills that reflect and advance critical thinking. The difference between mechanical and advanced writing is illustrated through a "Socratic Dialogue" with ChatGPT. Appropriate for classroom discussion at High School, College, Professional, and PhD levels.Item COVID-19 and Bone Loss: A Review of Risk Factors, Mechanisms, and Future Directions(Springer, 2024) Creecy, Amy; Awosanya, Olatundun D.; Harris, Alexander; Qiao, Xian; Ozanne, Marie; Toepp, Angela J.; Kacena, Melissa A.; McCune, Thomas; Orthopaedic Surgery, School of MedicinePurpose of review: SARS-CoV-2 drove the catastrophic global phenomenon of the COVID-19 pandemic resulting in a multitude of systemic health issues, including bone loss. The purpose of this review is to summarize recent findings related to bone loss and potential mechanisms. Recent findings: The early clinical evidence indicates an increase in vertebral fractures, hypocalcemia, vitamin D deficiencies, and a loss in BMD among COVID-19 patients. Additionally, lower BMD is associated with more severe SARS-CoV-2 infection. Preclinical models have shown bone loss and increased osteoclastogenesis. The bone loss associated with SARS-CoV-2 infection could be the result of many factors that directly affect the bone such as higher inflammation, activation of the NLRP3 inflammasome, recruitment of Th17 cells, the hypoxic environment, and changes in RANKL/OPG signaling. Additionally, SARS-CoV-2 infection can exert indirect effects on the skeleton, as mechanical unloading may occur with severe disease (e.g., bed rest) or with BMI loss and muscle wasting that has also been shown to occur with SARS-CoV-2 infection. Muscle wasting can also cause systemic issues that may influence the bone. Medications used to treat SARS-CoV-2 infection also have a negative effect on the bone. Lastly, SARS-CoV-2 infection may also worsen conditions such as diabetes and negatively affect kidney function, all of which could contribute to bone loss and increased fracture risk. SARS-CoV-2 can negatively affect the bone through multiple direct and indirect mechanisms. Future work will be needed to determine what patient populations are at risk of COVID-19-related increases in fracture risk, the mechanisms behind bone loss, and therapeutic options. This review article is part of a series of multiple manuscripts designed to determine the utility of using artificial intelligence for writing scientific reviews.