11351 research outputs found
Sort by
Deep Learning Models For Biomedical Data Analysis
Get PDFThe field of biomedical data analysis is a vibrant area of research dedicated to extracting valuable insights from a wide range of biomedical data sources, including biomedical images and genomics data. The emergence of deep learning, an artificial intelligence approach, presents significant prospects for enhancing biomedical data analysis and knowledge discovery. This dissertation focused on exploring innovative deep-learning methods for biomedical image processing and gene data analysis.
During the COVID-19 pandemic, biomedical imaging data, including CT scans and chest x-rays, played a pivotal role in identifying COVID-19 cases by categorizing patient chest x-ray outcomes as COVID-19-positive or negative. While supervised deep learning methods have effectively recognized COVID-19 patterns in chest x-ray datasets, the availability of annotated training data remains limited. To address this challenge, the thesis introduced a semi-supervised deep learning model named ssResNet, built upon the Residual Neural Network (ResNet) architecture. The model combines supervised and unsupervised paths, incorporating a weighted supervised loss function to manage data imbalance. The strategies to diminish prediction uncertainty in deep learning models for critical applications like medical image processing is explore. It achieves this through an ensemble deep learning model, integrating bagging deep learning and model calibration techniques. This ensemble model not only boosts biomedical image segmentation accuracy but also reduces prediction uncertainty, as validated on a comprehensive chest x-ray image segmentation dataset.
Furthermore, the thesis introduced an ensemble model integrating Proformer and ensemble learning methodologies. This model constructs multiple independent Proformers for predicting gene expression, their predictions are combined through weighted averaging to generate final predictions. Experimental outcomes underscore the efficacy of this ensemble model in enhancing prediction performance across various metrics.
In conclusion, this dissertation advances biomedical data analysis by harnessing the potential of deep learning techniques. It devises innovative approaches for processing biomedical images and gene data. By leveraging deep learning\u27s capabilities, this work paves the way for further progress in biomedical data analytics and its applications within clinical contexts.
Index Terms- biomedical data analysis, COVID-19, deep learning, ensemble learning, gene data analytics, medical image segmentation, prediction uncertainty, Proformer, Residual Neural Network (ResNet), semi-supervised learning
Piano Sonata no. 5 in C minor, op. 10, no. 1 Prestissimo
No full texthttps://digitalcommons.pvamu.edu/piano/1018/thumbnail.jp
A Study on Demographic Profile of Patients and Correlation Between Clinical and Radiological Findings in Oral and Oropharyngeal Malignancies
No full textOral and Oropharyngeal cancers are on rise year after year and Head and Neck Cancer rank sixth among all cancers worldwide. This study is an institutional based retrospective type of study conducted in a tertiary care hospital for a period of 6 months with an aim to identify the demographic profile of the patients and to correlate between the clinical and radiological findings of the disease using Cohen’s Kappa value. Both oral and oropharyngeal cancer was predominant in male population with commonest age being the 5th decade of life. Tongue was the most common site for oral cancer, on the other hand, base of tongue was commonest among oropharyngeal cancer. The disease extension was thoroughly examined clinically by palpation and 70 degree Hopkins Endoscope and also radiologically using appropriate imaging tools. The clinical and radiological T showed moderate degree of agreement for both oral and oropharyngeal cancers whereas clinical and radiological N showed a almost perfect degree of agreement
Real-World Image Restoration Using Degradation Adaptive Transformer-Based Adversarial Network
Get PDFMost existing learning-based image restoration methods heavily rely on paired degraded/non-degraded training datasets that are based on simplistic handcrafted degradation assumptions. These assumptions often involve a limited set of degradations, such as Gaussian blurs, noises, and bicubic downsampling. However, when these methods are applied to real-world images, there is a significant decrease in performance due to the discrepancy between synthetic and realistic degradation. Additionally, they lack the flexibility to adapt to unknown degradations in practical scenarios, which limits their generalizability to complex and unconstrained scenes.
To address the absence of image pairs, recent studies have proposed Generative Adversarial Network (GAN)-based unpaired methods. Nevertheless, unpaired learning models based on convolution operations encounter challenges in capturing long-range pixel dependencies in real-world images. This limitation stems from their reliance on convolution operations, which offer local connectivity and translation equivariance but struggle to capture global dependencies due to their limited receptive field.
To address these challenges, this dissertation proposed an innovative unpaired image restoration basic model along with an advanced model. The proposed basic model is the DA-CycleGAN model, which is based on the CycleGAN [1] neural network and specifically designed for blind real-world Single Image Super-Resolution (SISR). The DA-CycleGAN incorporates a degradation adaptive (DA) module to learn various real-world degradations (such as noise and blur patterns) in an unpaired manner, enabling strong flexible adaptation. Additionally, an advanced model called Trans-CycleGAN was designed, which integrated the Transformer architecture into CycleGAN to leverage its global connectivity. This combination allowed for image-to-image translation using CycleGAN [1] while enabling the Transformer to model global connectivity across long-range pixels. Extensive experiments conducted on realistic images demonstrate the superior performance of the proposed method in solving real-world image restoration problems, resulting in clearer and finer details.
Overall, this dissertation presents a novel unpaired image restoration basic model and an advanced model that effectively address the limitations of existing approaches. The proposed approach achieves significant advancements in handling real-world degradations and modeling long-range pixel dependencies, thereby offering substantial improvements in image restoration tasks.
Index Terms— Cross-domain translation, generative adversarial network, image restoration, super-resolution, transformer, unpaired training