# Research on the Precise Differentiation of Pathological Subtypes of Non-Small Cell Lung Cancer Based on 18F-FDG PET/CT Radiomics Features

## TL;DR

This study uses PET/CT imaging and clinical data to accurately distinguish between two types of lung cancer.

## Contribution

A novel nomogram model combining PET/CT radiomics and clinical features improves subtype differentiation in NSCLC.

## Key findings

- The nomogram model achieved an AUC of 0.880 and accuracy of 0.929 in validation.
- Combining PET/CT radiomics with clinical features outperformed models using only PET or CT data.
- The model showed high specificity (0.962) and sensitivity (0.808) in differentiating LUAD and LUSC.

## Abstract

Using 18F-FDG PET/CT radiomics features within and around tumors, combined with clinical characteristics, to accurately distinguish between different pathological subtypes of non-small-cell lung cancer (NSCLC). Radiomics feature extraction was performed on 18F-FDG PET/CT images of primary tumors and surrounding tumor regions using LIFE-x (5.2.0). Multivariate logistic regression analysis was used to construct a nomogram for distinguishing between lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). A nomogram model constructed by combining radiomic features extracted from 18F-FDG PET/CT images of tumors and surrounding tissues with clinical features can accurately distinguish between LUAD and LUSC.

Objectives: Employing 18F-FDG PET/CT radiomic properties both within and surrounding tumors, in conjunction with clinical attributes, to precisely differentiate among several pathological subtypes of non-small-cell lung cancer (NSCLC). Approaches: The study comprised 222 patients who received 18F-FDG PET/CT scans from January 2015 to December 2020 and were later diagnosed with NSCLC, encompassing 169 cases of lung adenocarcinoma (LUAD) and 53 cases of lung squamous cell carcinoma (LUSC). They were arbitrarily allocated into a training group and a validation group in a 7:3 ratio. Radiomics feature extraction was conducted on 18F-FDG PET/CT images of primary tumors and adjacent tumor regions with LIFE-x (5.2.0). A multivariate logistic regression analysis was employed to develop a nomogram for differentiating lung adenocarcinoma (LUAD) from lung squamous cell carcinoma (LUSC). The clinical efficacy of each model was assessed and contrasted utilizing accuracy (Acc), sensitivity (Sen), specificity (Spe), receiver operating characteristic curve (ROC), calibration curve, and decision curve analysis (DCA). Outcomes: The nomogram model that integrates 18F-FDG PET/CT radiomics features with clinical characteristics showed superior efficacy in differentiating adenocarcinoma from squamous cell carcinoma in NSCLC patients, surpassing models based only on PET or CT radiomics. The validation set exhibited an Area under curve (AUC) of 0.880, an Acc of 0.929, a Sen of 0.808, and a Spe of 0.962. This model exhibits the most superior overall performance in DCA. Conclusions: A nomogram model integrating radiomic features derived from 18F-FDG PET/CT images of tumors and adjacent tissues with clinical characteristics can effectively differentiate between LUAD and LUSC.