Case Study: Harvard Medical School reduces ultrasound annotation time from days to minutes with Encord

Harvard Medical School cuts vascular ultrasound annotation time from days to minutes with Encord

Harvard Medical School, along with Massachusetts General Hospital and Brigham and Women’s Hospital, faced a significant challenge in manually annotating Arterial Duplex Ultrasound (DUS) images for vascular research. This process was extremely time-consuming, prone to human error, and heavily dependent on scarce expert resources, which raised concerns about the reliability and reproducibility of their diagnostic measurements. To address this, they turned to Encord and its platform.

Using the Encord Annotate and Encord Active platforms, the researchers automated the image segmentation process. Encord enabled them to create models with minimal manual labeling, drastically cut annotation time, and evaluate model performance with key metrics. The solution reduced their annotation time from several days to just minutes. Encord also provided crucial insights, such as identifying that image blur negatively impacted model precision, leading to more reliable and efficient analyses.

Harvard Medical School