Low-grade magnetic resonance image enhancement using adaptive sigmoid transformation function

Medical images are crucial to monitor a patient’s health and provide an accurate diagnosis. Medical photos typically have various issues, including noise and poor contrast. A low-resolution magnetic resonance (MR) image is produced when there are irregularities in the structure, abnormalities in the pixel intensity allocation, and noise from the acquisition technique. Therefore, increasing contrast and reducing noise are required to improve the MRI image quality. There is no denying that the current approaches to quality improvement offend aesthetic sensibilities and have a significant disadvantage due to their high computational complexity and ineffectiveness.

This paper proposes a straightforward and computationally effective image enhancement framework to enhance visual quality and reduce distortions. An adaptive sigmoid transfer function (ASTF), derived from the sigmoid activation function of neural networks, is employed. The contrast-enhanced photos are created by combining ASTF with a modified histogram equalization method (ASTFMHE).

The application of the proposed method was assessed and contrasted using the performance measures Entropy, Patch-Based Contrast Quality Index (PCQI), Relative Contrast Measure (RCM), Modified Enhancement of Measure (MEME), and Edge Based Contrast Measure (EBCM). The suggested methodology, when compared to the other methods for Entropy (0.132), PCQI (1.085), RCM (0.169), MEME (75.852), and EBCM (15.831), yields the best findings for a pituitary tumor. Similarly, the proposed method gives best-suited results for glioma and meningioma, and there are no tumor MRI images except for entropy.

After considerable testing, it was discovered that the suggested method effectively increases visual contrast while retaining the original characteristics of the input picture and avoiding overly or inadequately improved images.