# Optimizing Layer Thickness in Multi-Planar Volume Reconstruction for Distinguishing Invasive Adenocarcinoma from Non-Invasive and Minimally Invasive Lesions in Pulmonary Nodules (≤15 mm): A Comparative Study with Conventional Lung Window Settings

## TL;DR

This study finds that a 10 mm layer thickness in multi-planar volume reconstruction improves accuracy in distinguishing invasive lung cancer from less severe lesions in small nodules.

## Contribution

The study identifies 10 mm as the optimal layer thickness for MPVR in diagnosing invasive adenocarcinoma in small pulmonary nodules.

## Key findings

- The 10 mm MPVR model achieved the highest AUC of 0.910 for differentiating invasive from non-invasive lesions.
- MPVR with 10 mm thickness outperformed conventional lung window settings in diagnostic accuracy.
- Performance metrics declined when layer thickness exceeded 10 mm.

## Abstract

Objective: To determine the optimal layer thickness for multi-planar volume reconstruction (MPVR) in differentiating invasive adenocarcinoma from non-invasive and minimally invasive lesions in pulmonary nodules (≤ 15 mm). Materials and Methods: This retrospective study enrolled a total of 601 solitary pulmonary nodules (≤15 mm) between June 2020 and February 2024, including 404 invasive adenocarcinomas (IAC), 80 micro-invasive adenocarcinomas (MIAs), 96 adenocarcinomas in situ (AISs), and 21 atypical adenomatous hyperplasias (AAHs). Thin-section computed tomography (TSCT) images with lung window settings and MPVR images with varying layer thicknesses (ranging from 2 to 14 mm with intervals of 2 mm) were analyzed for their morphological characteristics. Multivariate logistic regression analysis was employed to develop models for differentiating invasive adenocarcinoma from non-invasive and minimally invasive lesions. The model’s performances were further evaluated and compared to identify the optimal thickness for diagnosis. Results: The 10 mm MPVR model exhibited the best performance (AUC: 0.910, 95% CI [confidence interval]: 0.905–0.914; sensitivity: 0.906; specificity: 0.753; accuracy: 0.856; PPV: 0.883; and NPV: 0.796). As the MPVR layer thickness increased from 2 mm to 10 mm, model performance improved, with sensitivity rising from 0.870 to 0.906, specificity rising from 0.519 to 0.753, and accuracy increasing from 0.755 to 0.856. However, for layer thicknesses of 12 mm to 14 mm, all of them decreased. Furthermore, the overall performance of the 10 mm MPVR model surpassed that of the lung window model (AUC: 0.841, 95% CI: 0.831–0.844; sensitivity: 0.787; specificity: 0.760; accuracy: 0.778; PPV: 0.871; and NPV: 0.634). Conclusions: MPVR images with varying layer thicknesses can effectively distinguish invasive adenocarcinoma from non-invasive and minimally invasive lesions in pulmonary nodules ≤ 15 mm. Notably, the diagnostic performance of the 10 mm model was superior to model built with TSCT images, showing great potential as a precise and non-invasive tool for assessing the invasiveness of adenocarcinomas ≤ 15 mm.