Accelerating veterinary low field MRI acquisitions using the deep learning based denoising solution HawkAI.

Area of Science:

• Veterinary Medicine
• Medical Imaging
• Artificial Intelligence

Background:

• Magnetic Resonance Imaging (MRI) is crucial for veterinary diagnostics.
• Extended MRI scan times necessitate animal sedation and can impact image quality.
• Shorter scan times often lead to reduced image quality and diagnostic reliability.

Purpose of the Study:

• To evaluate the efficacy of HawkAI, a Generative Adversarial Network (GAN)-based denoising algorithm, in improving veterinary MRI image quality.
• To determine if faster MRI sequences processed by HawkAI are diagnostically comparable to Standard-Of-Care (SOC) sequences.
• To assess radiologist preference between HawkAI-processed faster MRI scans and traditional SOC scans.

Main Methods:

• Acquired Standard-Of-Care (SOC) and faster MRI sequences in veterinary patients.
• Applied the HawkAI denoising algorithm to the faster MRI sequences.
• Conducted a qualitative evaluation by radiologists using a Likert scale to compare HawkAI-processed images against SOC images.

Main Results:

• Radiologists showed a slight preference for HawkAI in Signal-to-Noise Ratio (SNR).
• No significant preference was observed for Artifacts Presence, Diagnosis Pertinence, or Lesion Conspicuity between HawkAI and SOC images.
• A very slight preference for SOC was noted in Spatial Resolution and Image Contrast.

Conclusions:

• HawkAI enables a significant reduction in veterinary MRI acquisition time (up to 2x faster).
• The algorithm effectively preserves diagnostic image quality, comparable to longer SOC scans.
• Faster MRI acquisition using HawkAI is feasible without major diagnostic drawbacks.