An AI-assisted navigation system reduced the number of needle passes during lung biopsies and the frequency of pneumothorax, a complication in which air enters the space around the lung. In a randomised trial involving 219 patients at three Chinese medical centres, GPS-Lung also shortened procedures and reduced the measured radiation dose-area product.[1]
The study, published in npj Digital Medicine, assigned 110 patients to GPS-Lung and 109 to standard cone-beam CT guidance. It tested the system during actual biopsy procedures, with clinicians using its navigation information to plan and position the needle.[1]
Making the route to a nodule clearer
Cone-beam CT provides three-dimensional imaging to help an operator reach a lung nodule and collect a tissue sample. GPS-Lung combines generative imaging with procedural navigation to support that task. The operator remains responsible for needle placement; the system provides guidance for a procedure that can be difficult when a lesion is small or awkwardly positioned.[1]
Outcomes in the 219-patient trial
| Outcome | GPS-Lung: 110 patients | Standard guidance: 109 patients |
|---|---|---|
| Mean needle passes | 2.0 | 2.5 |
| Mean procedure duration | 20.4 minutes | 23.2 minutes |
| Mean radiation dose-area product | 4,357.4 μGy·m² | 5,570.9 μGy·m² |
| Pneumothorax | 14 patients / 12.7% | 26 patients / 23.9% |
The authors report a mean reduction of 0.58 needle passes, with a 95% confidence interval of 0.18–0.97; the group means in the table are rounded. Mean procedure time fell by 2.8 minutes.[1]
A measured reduction in a clinical complication
Pneumothorax occurred in 14 patients in the GPS-Lung group and 26 in the control group. The published rates differ by 11.2 percentage points, with P = 0.033.[1]
The study did not show a statistically significant improvement in every complication measure, including haemorrhage and the assessed seven-day complications. Operators could not be blinded to the guidance system. Those details matter when interpreting the benefit alongside differences in operator experience, lesion characteristics and the equipment used at another hospital.[1]
The practical gain is a less burdensome route through an invasive procedure: fewer needle passes, a shorter procedure and lower exposure on the study’s radiation measure. The trial tests a defined intervention against guidance already used in practice.
For wider adoption, the next questions are implementation questions as well as clinical ones: how the system fits existing imaging equipment, what training operators need and whether the results carry over to other centres. The randomised comparison gives those decisions a stronger starting point than image-quality scores alone.
Sources & further reading
- A Generative AI-powered navigation system for precision CBCT-guided lung biopsies: a randomized controlled trial npj Digital Medicine · 21 Oct 2026 · Accessed 25 Oct 2026