Advancing AI Contouring at McGill with AutoContour v2.7

"Once you implement Zero-Click in AutoContour V2.7, it will become mission-critical within a few weeks."

At the McGill University Health Center in Montreal, AI contouring is no longer a novel addition to the clinic - it’s an embedded part of daily practice. Over the past several years, Dr Piotr Pater and his team have integrated automation into their workflow in a way that supports both efficiency and clinical consistency. Since adopting Limbus Contour in 2021, what began as a tool to accelerate contouring has become essential to daily operations. Today, it plays a central role in supporting both dosimetrists and radiation oncologists with consistent, high-quality results. As Piotr puts it. “AI auto-contouring has become mission-critical in our department.”

Now, as the team evaluates AutoContour v2.7, they are experiencing what feels less like a replacement and more like a natural evolution from Limbus Contour to a unified platform that preserves their existing workflows, while introducing additional flexibility.

One of the most immediate differences is the addition of a web-based interface. Where Limbus Contour workflows operated largely in the background, AutoContour v2.7 introduces greater visibility into workflows, giving users the ability to intervene when needed, especially in a small subset of cases where automation may need adjustment. As Piotr explains, “the web-based application allows us to monitor and easily resegment scans when needed. It gives end users control in the small percentage of cases where automation doesn’t capture everything.”

Importantly, this added control doesn’t come at the expense of automation. AutoContour v2.7’s Zero-Click workflow builds directly on the simplicity and automation that defined the team’s experience with Limbus Contour. “The beauty of Limbus Contour was its simplicity, automation, and good results,” says Piotr, “Zero-Click follows suit on all those aspects, with the added benefit of web-based control.” The balance that preserves what already works while expanding functionality has made the evaluation process seamless. Existing configurations from Limbus Contour can be carried forward into AutoContour v2.7. Allowing the team to maintain continuity without needing to rebuild workflows from scratch. “All our Limbus settings can be automatically imported, so the switch is not too difficult.”

Early evaluation suggests clear efficiency gains while maintaining the consistency the team relies on, particularly in more complex cases. “The end user now has more control through the web app, which reduces support work from our physics team.” Piotr also notes that it speeds up contouring for challenging cases, with “very similar contouring consistency between Limbus Contour and AutoContour v2.7.” Looking ahead, he adds, “As soon as clinically available, we will be switching to AutoContour 2.7 using the Zero-click workflow. This will be a seamless change for our clinic.”

The underlying model library also reflects this balance between continuity and expansion. He notes that many of the structure models in AutoContour v2.7 are carried over from Limbus Contour, providing continuity, while the addition of new models expands clinical capability. Following implementation, the team also plans to enhance their workflows further by incorporating additional structures into their templates and exploring the benefits of the AutoContour ESAPI scripting capabilities for deeper integration and automation.

Based on their experience so far, Piotr anticipates a smooth transition to full clinical adoption once the platform is widely available. For existing Limbus Contour users, he sees the transition as straightforward. “Just switch and appreciate the added flexibility and amazing AI contours,” he says.

For teams new to AI-driven contouring, he emphasizes the scale of impact, noting that adoption can rapidly transform departmental workflows. In his experience, “Once you implement Zero-Click, it will become mission-critical within a few weeks.”

Piotr highlights the merger of AutoContour and Limbus Contour as an important step forward in AI contouring. He notes that while Limbus helped establish efficient, automated workflows, AutoContour v2.7 brings these together with enhanced functionality and scale, supporting more comprehensive and adaptable clinical use.

Looking ahead, he believes the next major advances will come in areas such as deformable image registration, where clinical implementation has yet to fully match the pace of research. He sees strong potential for the combined Radformation team to drive a similar transformation, particularly through segmentation-guided approaches that support adaptive radiotherapy, including dose prescription and reporting.

Outside of the clinic, Piotr enjoys spending time with family, engaging in thoughtful discussions, exploring history through podcasts and museum visits, and relaxing weeks at a lakeside cottage.

Want to hear more from Dr. Pater?

Watch "Advancing AI Contouring with McGill University: AutoContour v2.7 Model Benchmarking and Clinical Evaluation" On Demand now.