Edge Preservation in Image Super-Resolution Using Transformer-Based Generative Models

Deep Feature Blend Attention: A New Frontier in Super Resolution Image Generation

Generative AI is an expanding domain that employs machine learning models to generate novel data that closely mimic pre existing data. ChatGPT and DALL-E can be customized for specific applications and are expected to transform healthcare, education, and communication. Generative Adversarial Networks (GANs) that can generate synthetic medical images closely mimicking actual patient data may substantially enhance machine learning model training datasets. They also translate medical images from one modality to another, improve medical imaging resolution, reduce radiation exposure, and boost image quality and detail. Despite their challenges, GANs have great potential in the field of medical imaging. The key obstacles are the need for Graphic Processing Units (GPUs) and computing resources to train GANs and the lack of established standards for generating synthetic images. Incorrectly labeled data for training other machine learning models can reduce performance, making ground-truth data labeling for healthcare AI more difficult. Generative AI is revolutionizing healthcare imaging, simplifying diagnosis, and propelling healthcare research and practice to new frontiers. Ensuring the reliability and safety of generated images in medical applications requires addressing ethical considerations and validating data.

For medical image analysis, there is always an immense need for rich details in an image. Typically, the diagnosis will be served best if the fine details in the image are retained and the image is available in high resolution. In medical imaging, acquiring high-resolution images is challenging and costly as it requires sophisticated and expensive instruments, trained human resources, and often causes operation delays. Deep learning based super resolution techniques can help us to extract rich details from a low-resolution image acquired using the existing devices. In this paper, we propose a new Generative Adversarial Network (GAN) based architecture for medical images, which maps low-resolution medical images to high-resolution images. The proposed architecture is divided into three steps. In the first step, we use a multi-path architecture to extract shallow features on multiple scales instead of single scale. In the second step, we use a ResNet34 architecture to extract deep features and upscale the features map by a factor of two. In the third step, we extract features of the upscaled version of the image using a residual connection-based mini-CNN and again upscale the feature map by a factor of two. The progressive upscaling overcomes the limitation for previous methods in generating true colors. Finally, we use a reconstruction convolutional layer to map back the upscaled features to a high-resolution image. Our addition of an extra loss term helps in overcoming large errors, thus, generating more realistic and smooth images. We evaluate the proposed architecture on four different medical image modalities: (1) the DRIVE and STARE datasets of retinal fundoscopy images, (2) the BraTS dataset of brain MRI, (3) the ISIC skin cancer dataset of dermoscopy images, and (4) the CAMUS dataset of cardiac ultrasound images. The proposed architecture achieves superior accuracy compared to other state-of-the-art super-resolution architectures.

In recent years, many fields have expanded their research methods through the integration of artificial intelligence. In the current medical field, it is widely used in image recognition to diagnose patient symptoms, train clinical prediction models, and improve clinical workflow processes to avoid missing treatment opportunities. These methods have all achieved good results so far. Now, generative AI is also impacting the development of the medical field, not only generating prescriptions that match patient conditions but also being applied in generating medical images for data augmentation and simulating various pathological conditions, thereby aiding in training deep learning models. Stroke is a leading cause of death or disability, ranking as the second leading cause of death worldwide, next to cancer. If cancers were categorized individually, stroke would surpass all single diseases, making it a primary focus of research in brain disease treatment. There are already studies where generative AI has been used to create medical images, which are then manually evaluated by doctors to determine if they were generated by AI. In this project, we plan to generate CT images of strokes using generative AI and use a trained model (to diagnose stroke) to improve accuracy. Through this project, we hope to solve the problem of scarce stroke CT images while also providing support for training stroke-related models.

Objective: Remote Sensing Image Captioning (RSIC) is a challenging research area that aims to automatically generate descriptive textual captions for Remote Sensing Images (RSIs) with recent advancements in deep learning, RSIC performance has significantly improved. Methods: This study presents a comparative analysis of two prominent approaches: Channel Attention Mechanism (CAM) and Generative AI (GenAI). The Channel Attention Mechanism emphasises on learning channel-wise with encoding-decoding, feature extraction and attention module. On the other hand, Generative AI, particularly models like GPT, leverages large-scale pretraining to diverse datasets for generates coherent and contextually relevant captions. It extracts vision-based feature and contains auto-tokenizer within the ViT Encoder-Decoder. Findings: This study evaluated on the RSICD dataset using metrics such as BLEU, METEOR, ROUGE-L, CIDEr, Precision, Recall, F1-Score, F2-Score, and Accuracy, GenAI outperforms CAM across all these metrics. Specifically, GenAI achieves a Precision of 0.90, Recall of 0.86, and Accuracy of 0.92. BLEU-4 scores for GenAI (0.7684) exceed CAM (0.4932), highlighting superior caption quality. However, both methods show limitations in accurately quantifying objects and specifying their coordinates within images. Novelty: The comparative analysis highlights the trade-off between traditional attention-based method and modern generative models. Keywords: Channel Attention Mechanism, Encoder-Decoder, Feature Extraction, Generative AI, Pre-Trained Models, RSICD Dataset

Secure IoT-enabled sharing of digital medical records: An integrated approach with reversible data hiding, symmetric cryptosystem, and IPFS

Image resolution enhancement is a problem of considerable interest in all medical imaging modalities. Unlike general purpose imaging or video processing, for a very long time, medical image resolution enhancement has been based on optimization of the imaging devices. Although some recent works purport to deal with image postprocessing, much remains to be done regarding medical image enhancement via postprocessing, especially in ultrasound imaging. We face a resolution improvement issue in the case of medical ultrasound imaging. We propose to investigate this problem using multidimensional autoregressive (AR) models. Noting that the estimation of the envelope of an ultrasound radio frequency (RF) signal is very similar to the estimation of classical Fourier-based power spectrum estimation, we theoretically show that a domain change and a multidimensional AR model can be used to achieve super-resolution in ultrasound imaging provided the order is estimated correctly. Here, this is done by means of a technique that simultaneously estimates the order and the parameters of a multidimensional model using relevant regression matrix factorization. Doing so, the proposed method specifically fits ultrasound imaging and provides an estimated envelope. Moreover, an expression that links the theoretical image resolution to both the image acquisition features (such as the point spread function) and a postprocessing feature (the AR model) order is derived. The overall contribution of this work is threefold. First, it allows for automatic resolution improvement. Through a simple model and without any specific manual algorithmic parameter tuning, as is used in common methods, the proposed technique simply and exclusively uses the ultrasound RF signal as input and provides the improved B-mode as output. Second, it allows for the a priori prediction of the improvement in resolution via the knowledge of the parametric model order before actual processing. Finally, to achieve the previous goal, while classical parametric methods would first estimate the model order and then the model parameters, our approach estimates the model parameters and the order simultaneously. The effectiveness of the methodology is validated using two-dimensional synthetic and in vivo data. We show that, compared to other techniques, our method provides better results from a qualitative and a quantitative viewpoint.

This chapter explores the transformative role of AI across various dimensions of healthcare, highlighting the interplay between analytical and generative AI in medical imaging, as well as the implications of generative AI in drug discovery and development. It reviews generative AI in the broader framework of P4 medicine—predictive, preventive, personalized, and participatory—and the potential of AI to improve health outcomes, while tackling the rising costs and increased demand for healthcare professionals in aging populations. The requirement for multimodal longitudinal datasets is examined, as well as the challenges around data sparsity and data bias, and the need for data equity. The chapter further evaluates inherent risks and the most relevant aspects of regulation relating to the use of AI in healthcare, personal data protection, and security. In conclusion, the chapter reviews the potential future impact of agentic AI, and technology convergence with robotics, zero-knowledge proofs and blockchain, and quantum computing.

Generative AI plays a pivotal role in medical imaging analysis, enabling precise diagnosis, treatment planning and disease monitoring. Techniques like generative adversarial networks (GANs) and variational autoencoders (VAEs) enhance medical imaging by generating synthetic images, improving reconstruction, segmentation and facilitating disease diagnosis and treatment planning. Nonetheless, ethical, legal and regulatory concerns arise regarding patient privacy, data protection and fairness. This paper offers an overview of generative AI in medical imaging analysis, highlighting applications, challenges and case studies. It compares results with traditional methods and examines potential implications on healthcare policies. The paper concludes with recommendations for responsible implementation and suggests future research and development directions.

While diffusion-based super-resolution (SR) methods have demonstrated promising results, they still face critical limitations in practical medical imaging applications. Recent methods focus on training on real-world image datasets, but these methods are prone to losing structural information when transferred to the medical imaging domain and suffer from extremely slow inference speed. Traditional diffusion-based super-resolution models in medical imaging face key challenges including image structure inconsistency, low computational efficiency, and unstable training. To address these issues, we propose FMSRdiff, an efficient latent diffusion framework that integrates consistency models with optical flow matching. Starting with latent space diffusion, we construct an efficient low-dimensional representation learning framework. This allows the diffusion process to be performed in a compressed latent space, significantly reducing computational complexity and memory requirements while maintaining high-quality reconstruction. We then design a two-stage training strategy that first learns stable noise prediction capabilities through optical flow matching, then switches to consistent model training to achieve a direct mapping from noisy to clean latent representations. This strategy not only ensures training stability but also significantly improves inference speed, reducing the 50-100 sampling steps required in traditional diffusion models to just one, significantly improving the model’s practicality. Furthermore, we introduce a hierarchical diffusion mechanism to achieve multi-scale feature processing from coarse to fine scales, effectively addressing the structural inconsistency problem in medical image super-resolution. Through cross-scale feature fusion and conditional guidance, the model better preserves the global structure and local details of the image, resulting in more natural and realistic super-resolution results. Experiments on the IXI and BRATS datasets demonstrate that our method achieves a 30% improvement in SSIM scores compared to existing state-of-the-art methods for medical imaging data. Through latent space diffusion, computational efficiency is increased by approximately 16 times, memory usage is significantly reduced.

Super-resolution imaging in gastrointestinal endoscopy is essential for enhancing diagnostic precision and reducing the risk of misdiagnosis. Although many researchers have introduced super-resolution methods for endoscopic images, these methods treat all image regions uniformly, neglecting the complex spatial dependencies among pixels. To address this problem, this paper proposes an adaptive convolutional feature fusion network for the super-resolution of gastrointestinal endoscopic images. The proposed model comprises two main modules: the adaptive multiscale feature extraction module (AMFEM) and the convolutional feature fusion block (CFFB). The AMFEM integrates several key components: (1) multiscale feature projection, which provides diverse receptive fields; (2) hierarchical dilated pooling, which maintains structural consistency; (3) channel-wise feature refinement, which selectively enhances channel representations, improving the discrimination of critical diagnostic features; and (4) feature-oriented spatial attention, which preserves key anatomical structures and highlights crucial regions. The CFFB further enhances the feature extraction by integrating depthwise-separable and pointwise convolutions, which capture fine spatial and channel-wise details while reducing complexity. Experiments were conducted on the Kvasir gastrointestinal endoscopic images dataset. The quantitative and qualitative results indicate that the proposed model outperforms state-of-the-art methods, achieving superior PSNR and SSIM scores while preserving clinically significant features essential for accurate diagnosis.

<p>Pneumonia is a severe lung infection, chest X-ray (CXR) image preferred to find infection. Real images lost its quality, resolution and other feature due to transmission. So good qualitative datasets are very limited. Quality enhancement in medical images is challenging task for researchers. And quality in clinical diagnosis of any disease in deep learning play a very important role. So, this paper presents an aspect with importance of quality in medical images CXR of a particular dataset and how to enhance and create new images with high quality resolution, that is re-used for classification in deep learning. Super resolution convolutional neural netwok (SRCNN) is deep learning based method, which is used for improving resolution in image. Super resolution means low resolution (LR) images from dataset is to be reconstructed or magnified into high resolution (HR). The objective behind this study is to measure the effect of super resolution with quality index, peak signal-to-noise ratio (PSNR), mean squared error (MSE), and structural similarity index measure (SSIM). This experinment performed on 200 images with 10 batches, each batch has 20 images from Kermany dataset, select LR images and converted into HR with SRCNN. Then we find PSNR value of image is increase upto 2 to 5 DB, and MSE of qood quality images is near to zero and MSE decrease up to 20-25, SSIM value have little variation due to same pattern is found in input and output images. Enhancement means highlight or improve the region of interest of pneumonic images. Main goal of this study is to preapare a modified dataset which is further used for classification.</p>

Reforestation is the primary and crucial solution towards the unapologetic destruction made by humans towards the natural ecological structure. Current methods are time-consuming and lack accurate spatial analysis. This paper introduces ReForGenAI, a novel system with Generative AI principles to automate the preliminary planning process and provide an actionable plan in hand. ReForGenAI accepts the longitude and latitude of a location as inputs. It leverages Google Earth Engine (GEE) to acquire the environmental data such as temperature, elevation, soil ph, soil type and climate type. A neural network trained on 480 tree species suggests the top suitable trees with a suitability score. Parallelly, it acquires the satellite and roadmap maps through Google Maps Static API for visual site context. Using computer vision techniques, the plantable area mask is identified, which excludes roads, buildings, waterbodies, and already existing trees. Plantable points are generated in the visual map using specific algorithms. A structured output provides a list of trees suitable for the location and a map that covers a radius of 175 meters with plantable tree points. This approach offers a quick, actionable plan using generative AI to address complex ecological planning tasks.

The convergence of Generative AI and agentic frameworks is fundamentally transforming enterprise data governance, offering a solution to the exponential growth in data complexity that traditional manual methods cannot address. This emerging paradigm shift introduces the concept of a Data Stewardship Co-Pilot—an intelligent, AI-powered partner that guides organizations through comprehensive data governance while democratizing access to complex data environments. The evolution from manual processes to intelligent automation employs specialized horizontal and vertical AI agents that autonomously handle diverse governance tasks with unprecedented precision and speed. These agents work within a sophisticated architecture comprising prompt stores, intuitive interfaces, vector-based indexing, context search engines, agent pools, knowledge graphs, and feedback mechanisms. By merging Generative AI's conversational capabilities with agentic frameworks' analytical depth, organizations can achieve significant improvements in efficiency, data quality, scalability, and collaboration while reducing operational costs and enhancing compliance standards. Leading technology platforms are actively facilitating these innovations, positioning AI-assisted data stewardship as a strategic differentiator in the increasingly data-driven business landscape.

High-resolution construction images are essential for quality monitoring and progress management. Addressing the challenge that existing super-resolution models often struggle to balance reconstruction quality with real-time inference speed in complex construction environments, this paper proposes a super-resolution reconstruction model based on attention mechanisms and Generative Adversarial Networks (GANs). The generator captures fine details through a Multi-scale Shallow Feature Extraction (MSFE) module. The deep feature mapping module introduces the SORRDB (SFT-Octave-Residual in Residual Dense Block) as its fundamental unit. This module integrates SFT layers for dynamic spatial feature modulation and incorporates Octave Convolution to decouple high- and low-frequency information, thereby reducing computational redundancy, while scSE attention mechanisms further enhance high-level structural features. The discriminator integrates Swin Transformer to capture both local and global features, enhancing the quality and realism of generated images. Experimental results on the SODA dataset demonstrate that the model achieves superior performance with PSNR, SSIM, and LPIPS scores of 28.063 dB, 0.825, and 0.197, respectively. Crucially, the model achieves an inference speed of 32.0 FPS on an NVIDIA GeForce RTX 4070, meeting the requirements for real-time monitoring. Therefore, the proposed model not only significantly enhances the perceptual quality and structural integrity of construction images but also achieves an optimal balance between performance and efficiency, offering substantial benefits for intelligent construction monitoring.