Scribd
Upload
43 views2 pages

Medical Iqe Updated

This paper presents a task-driven image quality enhancement (IQE) strategy called GradProm, which improves the quality of medical images by aligning the gradients of image enhancement and visual recognition models. The approach is designed to enhance the perceived quality of images for human evaluation, thereby improving recognition accuracy in medical imaging tasks. GradProm is versatile and applicable to various medical image modalities, with potential extensions to other medical image analysis tasks.

Uploaded by

mr.siva.reddy.666
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
43 views2 pages

Medical Iqe Updated

This paper presents a task-driven image quality enhancement (IQE) strategy called GradProm, which improves the quality of medical images by aligning the gradients of image enhancement and visual recognition models. The approach is designed to enhance the perceived quality of images for human evaluation, thereby improving recognition accuracy in medical imaging tasks. GradProm is versatile and applicable to various medical image modalities, with potential extensions to other medical image analysis tasks.

Uploaded by

mr.siva.reddy.666
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

TASK-DRIVEN IMAGE QUALITY ENHANCEMENT IN MEDICAL

IMAGING

### ABSTRACT
Thanks to recent achievements in task-driven image quality enhancement (IQE) models like ESTR
(2023), the image enhancement model and visual recognition model can mutually enhance each
other's quantization while producing high-quality processed images that are perceivable by human
vision systems. Different levels of vision tasks have varying and sometimes conflicting requirements
for image features. To address this problem, this paper proposes a generalized gradient promotion
(GradProm) training strategy...

### INTRODUCTION
Medical imaging plays an increasingly important role in modern medicine, enabling physicians to
visualize internal anatomical structures and evaluate disease progression. Image quality significantly
impacts medical image analysis (MIA), as higher-quality images are expected to yield more accurate
recognition performance. Consequently, various medical image quality enhancement (IQE)
approaches have been developed, such as super-resolution and image denoising. The input of
these approaches is typically a l...

"Human evaluation" is an important practical criterion for medical IQE tasks, as the primary goal is to
improve the perceived image quality for human visual systems. This paradigm has been shown to
enhance recognition accuracy in both upstream and downstream tasks. Our experiments involved
publicly available synthetic and real medical datasets from diverse image modalities, including ISIC
2018, COVID-CT, Lizard, and CAMUS.

Medical image quality enhancement is a critical area of image processing aimed at improving the
visual appearance and diagnostic accuracy of medical images such as X-rays and tomography
scans. Multi-task learning (MTL) is a popular approach in computer vision that leverages useful
knowledge from related tasks to improve overall performance. We observe that, compared to
scenarios where training and testing images come from the same dataset, both ESTR and
GradProm exhibit relatively lower performance when ...

### CONCLUSION
In this paper, we proposed a versatile and effective training strategy, GradProm, for medical IQE
within a task-driven framework. Applicable to a broad range of image modalities, GradProm is not
restricted to a specific medical image modality. The GradProm strategy updates parameters only
when the gradients of the two sub-models align in the same direction, as determined by their cosine
similarity. As a generalized method, GradProm can be extended to other task-driven training
processes, such as multi-ob...

Additionally, we plan to investigate the application of GradProm to other medical image analysis
tasks, including medical image registration and reconstruction.

You might also like