Clinical Evaluation of the performance of Mediq RT compared to other commercial solutions

At the end of 2023, Synaptiq underwent a clinical evaluation assessing the performance of its AI-based auto-contouring solution, Mediq RT, in comparison to other commercially available solutions. The clinical evaluation was performed by the German Oncology Center on a set of 80 patients (20 breast, 20 head and neck, 20 lung, 20 prostate).

In modern radiotherapy, contouring is one of the biggest workflow bottlenecks and one of the biggest sources of variability. A benchmarking study published in Frontiers in Oncology (2023, Paul Doolan) expanded its comparison to include Synaptiq and evaluated six commercial systems using the same dataset and the same metrics. The takeaway is simple: AI can standardize contour quality and cut editing time massively, which matters now because oncology teams are overloaded and case volumes keep rising.

About German Oncology Center

What AI-based auto contouring solutions were compared?

How were the AI auto-contouring solutions compared?

Contours were evaluated with geometric measures that act as today’s practical “gold standard” for segmentation benchmarking:

• Volumetric Dice (vDSC): Overlap between two volumes (0 to 1). Great for large organs, can look “better” on big structures.
• Surface Dice (sDSC, tolerance = 0): Surface agreement withno tolerance, so boundary placement errors show up fast.
• Hausdorff Distance (HD): Worst-case boundary mismatch. Lower is better, highlights extreme outliers.
• Added Path Length (APL): How much contour line needs to be changed to match reference. Very close to “editing effort.”

Results of the clinical evaluation of Mediq RT

Synaptiq’s Mediq RT was benchmarked against five established commercial AI auto-contouring solutions using identical datasets and evaluation methods. The assessment focused on geometric accuracy and time-saving potential across four anatomical sites: breast, head and neck, lung, and prostate.

How accurate is Mediq RT in practice?

Across all four anatomical sites (breast, head and neck, lung, prostate), Synaptiq’s solution achieved high geometric similarity scores in volumetric and surface Dice metrics, generally comparable to or slightly below the top-performing systems depending on the structure. Median vDSC values for large, well-defined organs such as the lungs, liver, and bladder were consistently high (often above 0.94), while more complex or poorly visible structures (e.g., oesophagus, small glands) showed moderate scores similar to other vendors.

In terms of Hausdorff Distance (HD), Synaptiq’s results indicated competitive boundary accuracy, with lower HD values for most structures, especially in the lung and head-and-neck datasets, where extreme outliers were minimal. The Added Path Length (APL) results reflected moderate correction requirements, aligning with clinical observations that its contours required less extensive editing for larger, easily segmented organs, but more adjustments for small, complex structures.

From a workflow efficiency perspective, Synaptiq demonstrated substantial time savings over manual contouring. Average correction times were 5.9 minutes for breast cases, 2.7 minutes for head and neck, 3.5 minutes for lung, and 1.8 minutes for prostate. This translated into relative time savings of 73% for breast, 97,2% for head and neck, 87% for lung, and 95.7% for prostate, comparable to or exceeding other systems in certain categories.

Overall, the evaluation concluded that Mediq RT delivers clinically acceptable contours with significant time-saving potential, performing on par with other established commercial AI solutions and particularly excelling in workflow efficiency for complex head-and-neck cases.

13.5x faster than manual contouring 

From a workflow efficiency perspective, Synaptiq demonstrated substantial time savings over manual contouring. Average correction times were 5.9 minutes for breast cases, 2.7 minutes for head and neck, 3.5 minutes for lung, and 1.8 minutes for prostate. This translated into important time saving for all regions and with an astonishing 97,2% time reduction for head and neck comparable to or exceeding other systems in certain categories.

1st place for Head&Neck

Table 1. Median volumetric Dice (vDSC) and median surface Dice (sDSC, with no additional tolerance) similarity coefficients for twenty head and neck patients.

Figure 1 Distribution of vDSC for head and neck patients.

What does this clinical evaluation mean to doctors who use Mediq RT?

This clinical evaluation is an independent validation that Mediq RT performs at the level of established commercial systems, using the same dataset, the same methodology, and the same evaluation metrics.

The most important results:

• 6% average time savingacross breast, head & neck, lung, and prostate
• ~13.5× fastercontouring compared to manual work
• #1 overall in Head & Neck, the most complex and time-consuming site

For clinicians, the impact is practical. Less time spent correcting routine contours means more time available for plan review, peer discussion, and quality checks, where clinical judgment actually changes outcomes.

For Synaptiq, being measured head-to-head against five international vendors and coming out with top performance in Head & Neck, plus strong efficiency across all sites, positions Mediq RT as a proven solution for large-scale clinical deployment.