Case Study: Duke University accelerates clinical-grade ultrasound imaging research with Google for Education

Duke University achieves 0.940 SSIM and cuts ultrasound imaging results time from 3 weeks to 1 day with Google for Education

Duke University researchers faced a challenge in medical imaging, as the proprietary post-processing used to create clinical-grade ultrasound images varies across manufacturers. This made it difficult to establish baselines and hindered clinical translation research. To solve this, they turned to Google for Education, utilizing Google Cloud products like Compute Engine and Colab to develop a universal open-source framework.

Using Google Compute Engine for batch training, the team developed a deep-learning tool called MimickNet. The solution from Google for Education dramatically accelerated their research, reducing the time for experiments from three weeks to just one day. The tool successfully mimicked clinical-grade scanners, achieving a near-perfect image similarity score, and its open-source model is now easily accessible to other researchers via Google Colab.

Duke University