Abstract Rationale Early diagnosis of lung cancer is critical for patient survival and relies on a successful initial biopsy. Depending on lesion location and size, current diagnostic yield can be as low as 42%. At early stages, more than 70% of cancer nodules are located in difficult-to-reach regions, making early diagnosis challenging. Here, we evaluated whether a smart stylet equipped with an impedance-based sensor can provide in situ information on lung lesions (tool-in-lesion data) during a bronchoscopic procedure immediately prior to biopsy tissue sampling. The goal is to reduce uncertainty in confirming the target lesion and enhance diagnostic yield. Methods First-in-human, single-arm study, conducted in Australia and France involving subjects with central or peripheral lung lesions eligible for transbronchial biopsy. In 26 patients, impedance data were acquired and annotated during image-guided bronchoscopic lung biopsy in healthy tissues and lesions (cancer, inflammation, necrosis, fibrosis), from which biopsy samples were collected for histopathological analysis. The 0.36mm diameter smart stylet was placed inside the biopsy needle and readings were taken immediately prior to biopsy. In situ impedance data were complemented by a previously collected, ex vivo dataset for the development and validation of prediction models designed to differentiate lesions from healthy lung tissue, and to distinguish cancerous tissue from all other tissues. A patient cross-validation analysis was used to evaluate prediction models’ performance as per accuracy, sensitivity, and specificity, in which impedance acquisitions within lesions or cancerous tissues were defined as the positive class, whereas other tissue measurements were treated as the negative class. Results Impedance data was successfully obtained from the smart stylet within lesions in all patients, with no adverse events. The prediction model differentiating healthy from lesion tissues, demonstrated accuracy of 80.9%, sensitivity of 88.5%, and specificity of 71.4%. The prediction model differentiating cancer from all other tissues, including necrosis, demonstrated an accuracy of 78.7%, sensitivity of 78.3%, and specificity of 79.2%. Analysis of these initial results revealed the impedance data prediction models’ learning curves are in a linear growth phase (Figure 1 A&B). Conclusions Impedance-based prediction models can accurately identify lesions or cancer amid the complex conditions of in situ lung tissue. With learning curves suggesting the potential to exceed 90% overall performance, these models show clear promise for real-time decision support during lung biopsy. By confirming relevant sampling sites and tool-in-lesion, impedance-based technology could significantly boost biopsy diagnostic yield and shorten the path to lung cancer diagnosis and treatment. This abstract is funded by: Sensome
Hanna et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: