Supplementary Material for: Learning shape sensing robotic assisted bronchoscopy after mastering advanced image guided navigation bronchoscopy

Introduction
Shape sensing robotic-assisted bronchoscopy (ssRAB) is a novel technique for the diagnosis of peripheral pulmonary lesions (PPL). There are no prior studies who have assessed the learning curve using learning curve cumulative summation (LC-CUSUM) and cumulative summation (CUSUM) analyses with diagnostic yield as the main endpoint.
Methods
We performed a single-center analysis of the learning curve of ssRAB combined with CBCT procedures for the diagnosis of PPL performed by two bronchoscopist experienced with 3D-imaging-guided bronchoscopy techniques using 1) an LC-CUSUM followed by CUSUM analysis, 2) a CUSUM analysis, with strict diagnostic yield as endpoint. We will compare these methods.
Results
A total of 131 patients with a median lesion size of 12 millimeter (9-18mm) were navigated by two bronchoscopists. In the first method, the LC-CUSUM analysis indicated that both bronchoscopists were statistically declared proficient after 43 and 42 procedures, with subsequent CUSUM analysis confirming sustained performance thereafter. In the second method, CUSUM analysis revealed that both bronchoscopists were deemed in control during all procedures.
Conclusion
The CBCT enhanced ssRAB performance of bronchoscopists experienced with CBCT guided navigation bronchoscopy was in control from the start. LC-CUSUM and CUSUM are useful tools for assessing the learning curve and procedural performance. While CUSUM indicated that performance of the experienced bronchoscopists were in control from the start, LC-CUSUM inherently assumes initial non-proficiency leading to a larger number of procedures required to establish proficiency with statistical certainty. In the diagnosis of PPL, there will always be variability in performance, which may be attributable to lesion- or patient-specific characteristics.