2023 World Conference on Lung Cancer (Posters)-EP07.02. The Prediction of Pathological Pleural Invasion of Lung Cancer by Artificial Intelligence Image Analysis

EP07.02. The Prediction of Pathological Pleural Invasion of Lung Cancer by Artificial Intelligence Image Analysis - PDF(Abstract)

Back to course

Pdf Summary

This presentation at the WCLC 2023 conference discussed the use of artificial intelligence (AI) image analysis to predict pathological pleural invasion in patients with early-stage non-small cell lung cancer. The study was conducted at Tokyo Medical University Hospital and included 556 patients who underwent surgical resection for lung cancer between 2011 and 2018.<br /><br />The researchers performed AI analysis on preoperative CT images of the patients using Synapse Vincent software. This analysis provided radiological findings of the lung tumors, including 3D imaging and 22 radiological features with confidence scores. The patients were divided into training and test cohorts, and the association between these features and pleural invasion was statistically assessed.<br /><br />The results showed that pleural invasion was present in 32% of the patients. Univariable analysis identified several radiological features, such as clear boundary, bronchus translucency, pleural contact, solid nodule, part-solid nodule, and grand grass nodule, that were associated with pleural invasion. However, binary logistic regression analysis using stability selection revealed that solid nodule and pleural contact were the most significant factors related to pleural invasion.<br /><br />The researchers conducted receiver operating characteristic (ROC) analysis to evaluate the predictive ability of these factors. The area under the curve (AUC) was 0.831 in the training dataset and 0.782 in the test dataset, indicating that the AI analysis accurately predicted pleural invasion. The sensitivity and specificity for pleural invasion were calculated to be 0.739 and 0.657, respectively, in the test dataset.<br /><br />In conclusion, this study demonstrated that AI analysis can accurately predict pathological pleural invasion in patients with early-stage non-small cell lung cancer. The researchers suggested that integrating AI software analysis into the preoperative assessment could lead to more precise evaluations of pleural invasion.

Asset Subtitle

Wakako Nagase

Meta Tag

Speaker Wakako Nagase

Topic Early-Stage NSCLC: New Technology & Innovations

Keywords

artificial intelligence

image analysis

pathological pleural invasion

non-small cell lung cancer

Tokyo Medical University Hospital

preoperative CT images

radiological features

solid nodule

binary logistic regression analysis

predictive ability