Scholarlink

The aim of this research was to recommend the best source quality of pictures for T1- and T2-weighted magnetic resonance imaging (MRI) pictures that use conditional generative adversarial networks (GAN). An aggregate of 2,024 pictures from 104 patients’ scans between 2017 and 2018 were utilized. GAN was used to develop the predicting frameworks for T1-weighted to T2-weighted MRI pictures and T2-weighted to T1-weighted MRI images. For the supplied pictures two different gray scale level converting techniques (simple and adaptable) and two different picture sizes (512 512 and 256 256) were considered. Using a straightforward converting technique, the images were split into 256 levels to get between 16-bit to 8-bit. In order to employ the adaptable conversion approach, the unwanted levels in 16-bit images were removed before they were transformed to 8-bit pictures by splitting them into the result of multiplying the highest pixel value by 256. Using an adaptable conversion approach, the mean absolute error (MAE) as 0.15 for T1-weighted to T2-weighted MRI pictures and 0.17 for T2-weighted to T1-weighted MRI images, which was considered the minimum. Additionally, the adapted conversion technique has the highest peak signal-to-noise ratio (PSNR), as well as the lowest mean square error (MSE) with root mean square error (RMSE). The dimension of the picture affected the duration of computation. The precision of forecast is influenced by the picture size and input resolution. Despite the necessity for drawn-out checks, the suggested theory and technique of the projection architecture can aid in increasing the adaptability and caliber of multi-contrast MRI tests.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Approach for Predicting Accuracy in Image to Image Translation by Means of Generative Adversarial Networks (GAN)

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vankudothu Malsoru,
  • K. Shilpa,
  • Avala Raji Reddy,
  • Abdul Subhani Shaik,
  • D. Sudha,
  • B. Archana

摘要

The aim of this research was to recommend the best source quality of pictures for T1- and T2-weighted magnetic resonance imaging (MRI) pictures that use conditional generative adversarial networks (GAN). An aggregate of 2,024 pictures from 104 patients’ scans between 2017 and 2018 were utilized. GAN was used to develop the predicting frameworks for T1-weighted to T2-weighted MRI pictures and T2-weighted to T1-weighted MRI images. For the supplied pictures two different gray scale level converting techniques (simple and adaptable) and two different picture sizes (512 512 and 256 256) were considered. Using a straightforward converting technique, the images were split into 256 levels to get between 16-bit to 8-bit. In order to employ the adaptable conversion approach, the unwanted levels in 16-bit images were removed before they were transformed to 8-bit pictures by splitting them into the result of multiplying the highest pixel value by 256. Using an adaptable conversion approach, the mean absolute error (MAE) as 0.15 for T1-weighted to T2-weighted MRI pictures and 0.17 for T2-weighted to T1-weighted MRI images, which was considered the minimum. Additionally, the adapted conversion technique has the highest peak signal-to-noise ratio (PSNR), as well as the lowest mean square error (MSE) with root mean square error (RMSE). The dimension of the picture affected the duration of computation. The precision of forecast is influenced by the picture size and input resolution. Despite the necessity for drawn-out checks, the suggested theory and technique of the projection architecture can aid in increasing the adaptability and caliber of multi-contrast MRI tests.