Pituitary neuroendocrine tumor: evaluation with super resolution deep learning reconstruction : Research

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Pituitary neuroendocrine tumors (PitNET) require high-quality MRI for accurate diagnosis.
• Conventional MRI reconstruction techniques like zero-filling interpolation (ZIP) may limit image quality and diagnostic confidence.
• Super-resolution deep learning reconstruction (SR-DLR) is a novel technique with potential to enhance medical image resolution and clarity.

Purpose of the Study:

• To evaluate the impact of SR-DLR on pituitary MRI image quality.
• To assess the effect of SR-DLR on the evaluation of pituitary neuroendocrine tumors (PitNET).
• To compare SR-DLR with conventional ZIP technique for pituitary MRI.

Main Methods:

• Retrospective analysis of 29 PitNET patients' pituitary MRI scans.
• Reconstruction of T2-weighted coronal images using SR-DLR and ZIP.
• Qualitative assessment by three readers for image quality and diagnostic acceptability.
• Quantitative analysis including signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), edge rise slope (ERS), and full width at half maximum (FWHM).

Main Results:

• SR-DLR demonstrated superior qualitative image analysis scores compared to ZIP across all evaluated items (p < 0.001).
• Significantly higher SNR and CNR were observed with SR-DLR versus ZIP (p < 0.001).
• SR-DLR showed enhanced spatial resolution, indicated by improved ERS and reduced FWHM values compared to ZIP.

Conclusions:

• SR-DLR significantly improves the image quality of pituitary MRI compared to conventional ZIP.
• SR-DLR tends to enhance inter-reader agreement in assessing pituitary stalk deviation.
• The SR-DLR algorithm shows promise for improving diagnostic accuracy in PitNET imaging.