How AI Code Assistant Baidu Comate Boosted Medical Imaging Processing by 9×

Graduate student Wang Rongsheng and his lab at Macau University of Science and Technology faced a routine but time‑consuming task: converting large volumes of DICOM medical images (about 150 GB per batch) into PNG format and generating accompanying JSON metadata, a process that previously required three people working a full week.

In late 2022 they adopted Baidu Comate, an AI code assistant built on Baidu’s Wenxin large‑language model and integrated with major IDEs such as VS Code, JetBrains, and Xcode. Comate offers real‑time code continuation, natural‑language‑driven code generation, annotation generation, and a private‑knowledge‑enhancement module.

Using Comate’s “real‑time continuation” and “code‑from‑prompt” features, the team could generate complete functions, loops, and conditionals simply by describing the desired behavior. The assistant also produced inline comments automatically, improving code readability and team collaboration.

For example, the following natural‑language instruction was fed to Comate to create a full batch‑processing pipeline:

Use Python to read all medical images in the "image" folder, split each image into 512×512 overlapping windows, record window metadata in a JSON file that can reconstruct the original image, and apply the same windowing to the "label" folder, saving results to "image_output" and "label_output".

With these capabilities, the lab reduced the processing time from seven days (three people) to two days for a single developer, a productivity increase of more than nine times.

Beyond speed, Comate’s “code‑generation‑annotation” function allowed the team to add missing comments with a single click, standardising documentation across contributors. The newly released “Comate Open Platform” let Wang import his personal “code repository” of 60‑70 reusable modules, enabling private‑knowledge enhancement that improved code‑generation accuracy by 10‑20% and eliminated manual code‑search effort.

Wang, who recently completed his master’s degree and plans to pursue a PhD, expects AI tools like Comate to become integral to interdisciplinary research, allowing developers to focus on high‑level design while the AI handles repetitive coding tasks.