An international life sciences company trains an AI tool to classify millions of clinical particle images for medical research.


Our client has collected and stored tens of millions of urinalysis particle images that need human classification in order to delineate each element/particle present. For the first phase of the proof of concept (POC), our client needed a partner who could classify the particle images to train the AI. With the fully classified data, they will train the artificial intelligence device to take over this process. There were approximately 68,000 total images to be classified over the course of the POC.

Exceeded production goal by training AI to classify 4.6 images per minute on average


Apex provided a managed medical technician workforce including an engagement manager to interface with the client’s leadership. Apex leveraged our workforce management processes including demand planning, consultant management, onboarding, and retention for the client. Our engagement manager submitted weekly status reports and held a steady meeting cadence with the client's project leader. Based on the number of images to complete and the assumed pace to label the images, we leveraged four medical technicians in order to properly balance the workload. Apex oversaw weekly classification progress and monitored productivity to balance images labeled against the completion timeline.


We were able to classify 71,298 images, exceeding the target of 68,000, on time and under budget. For phase one of the POC we were able to identify best practices, target pace of images classified, accuracy targets, and productivity targets of medical technicians to work for future phases. As a best practice, we provided feedback on the client’s portal which was developed to aid in this process. Due to our success, the client has agreed to incorporate us into additional phases of the POC and has referred Apex to other parts of the organization for similar efforts.