The Machine Learning Model for Distinguishing Pathological Subtypes of Non-Small Cell Lung Cancer

Abstract

PurposeMachine learning models were developed and validated to identify lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) using clinical factors, laboratory metrics, and 2-deoxy-2[18F]fluoro-D-glucose ([18F]F-FDG) positron emission tomography (PET)/computed tomography (CT) radiomic features.MethodsOne hundred and twenty non-small cell lung cancer (NSCLC) patients (62 LUAD and 58 LUSC) were analyzed retrospectively and randomized into a training group (n = 85) and validation group (n = 35). A total of 99 feature parameters—four clinical factors, four laboratory indicators, and 91 [18F]F-FDG PET/CT radiomic features—were used for data analysis and model construction. The Boruta algorithm was used to screen the features. The retained minimum optimal feature subset was input into ten machine learning to construct a classifier for distinguishing between LUAD and LUSC. Univariate and multivariate analyses were used to identify the independent risk factors of the NSCLC subtype and constructed the Clinical model. Finally, the area under the receiver operating characteristic curve (AUC) values, sensitivity, specificity, and accuracy (ACC) was used to validate the machine learning model with the best performance effect and Clinical model in the validation group, and the DeLong test was used to compare the model performance.ResultsBoruta algorithm selected the optimal subset consisting of 13 features, including two clinical features, two laboratory indicators, and nine PEF/CT radiomic features. The Random Forest (RF) model and Support Vector Machine (SVM) model in the training group showed the best performance. Gender (P=0.018) and smoking status (P=0.011) construct the Clinical model. In the validation group, the SVM model (AUC: 0.876, ACC: 0.800) and RF model (AUC: 0.863, ACC: 0.800) performed well, while Clinical model (AUC:0.712, ACC: 0.686) performed moderately. There was no significant difference between the RF and Clinical models, but the SVM model was significantly better than the Clinical model. ConclusionsThe proposed SVM and RF models successfully identified LUAD and LUSC. The results indicate that the proposed model is an accurate and noninvasive predictive tool that can assist clinical decision-making, especially for patients who cannot have biopsies or where a biopsy fails.

The role of artificial intelligence based on PET/CT radiomics in NSCLC: Disease management, opportunities, and challenges

“…The gold standard of histological classification is pathological evaluation. At present, many techniques can be used for tissue diagnosis and the preferred choice is image-guided needle core biopsy ( 38 ). Unfortunately, as an invasive examination, the operational risks of biopsy are inevitable, including a series of complications such as infection.…”

“…Zhao et al. use the Boruta algorithm to determine an optimal subset of 13 features, including two clinical features(sex and smoking history), two laboratory indicators(CEA and SCCA), and nine PEF/CT radiomic features ( 38 ). The clinical features they included were the same as those of Ren et al ( 44 ), but there were differences in laboratory indicators.…”

“…Moreover, the spatio-temporal pathological heterogeneity of tumors limits the ability to capture biodiversity or disease evolution in a single biopsy (39). By contrast, the extraction of quantitative radiological features from noninvasive imaging can reflect a wealth of information about genotype, tumor microenvironment, and sensitivity to treatment (38). The clinical features they included were the same as those of Ren et al (44), but there were differences in laboratory indicators.…”

Risk prediction of gestational diabetes mellitus with four machine learning models

“…Unlike the results of the LR model, the results of the RF model showed that clinical biochemical indicators are generally better predictors of risk than general demographic characteristics and previous pregnancy history. This is probably because the variable lter of the RF model is not sensitive to indicators that appear less frequently in the dataset [55] . It is di cult for logistic regression model to apply to collinear variables and it has poor prediction ability for high-dimensional, multi-sample, non-linear data.…”

Quantitative Analysis of TP53-Related Lung Cancer Based on Radiomics

“… 21 , 22 Associating image features with tumor gene–protein features or tumor phenotypes can develop models for cancer diagnosis, patient prognosis, or relative tumor heterogeneity, especially for patients who are unable to perform biopsy or whose biopsy fails. 23 In the long run, radiomics will not replace the role of radiologists. Instead, it will be a powerful tool for radiologists.…”