JMIR Publications

ABSTRACT

Background:

During the diagnosis and treatment of non-small cell lung cancer (NSCLC), detecting the risk of its recurrence in an early phase is still challenging. Recent studies have investigated the radiomics-based machine learning (ML) models for detecting the risk of recurrence in NSCLC. However, there is still insufficient systematic evidence to prove its efficiency.

Objective:

Under this background, this research intends to ascertain the efficiency of radiomics-based ML in forecasting the risk of recurrence in NSCLC at an early stage to provide evidence-based support for its clinical application.

Methods:

For acquiring research on radiomics-based models for forecasting the risk of recurrence in NSCLC, Cochrane Library, Web of Science, PubMed, and Embase were systematically retrieved, up to October 24th, 2025. The Radiomics Quality Score (RQS) was employed to appraise the eligible studies. Subgroup analyses were conducted according to variables of model, background of treatment, stage of lung cancer, and pathological type.

Results:

Ultimately, 26 eligible studies in total were included, covering 7,964 patients with NSCLC. According to the meta-analysis, the c-index of radiomics-based ML models for forecasting the risk of recurrence in NSCLC was 0.849 (95% CI: 0.834-0.865, 95% PI: 0.633-0.998) in the training set. Specifically, the pooled c-index was 0.776 (95% CI: 0.853-0.900) among the patients receiving the stereotactic body radiation therapy (SBRT) and 0.827 (95% CI: 0.807-0.848) among those who received surgeries combined with other adjuvant treatment regimens. The c-index of the radiomics-based ML models combined with clinical features for forecasting the risk of recurrence in NSCLC was 0.835 (95% CI: 0.824-0.846, 95% PI: 0.721-0.943) in the training set. In contrast, the c-index of radiomics-based ML models for forecasting the risk of recurrence in NSCLC was 0.856 (95% CI: 0.837-0.876, 95% PI: 0.692-0.967) in the validation set. Specifically, among patients receiving SBRT, the pooled c-index was 0.885 (95% CI: 0.857-0.914), and among those who received surgeries combined with other adjuvant treatment regimens, the pooled c-index was 0.825 (95% CI: 0.799-0.853). The c-index of radiomics-based ML models combined with clinical features for forecasting the risk of recurrence in NSCLC was 0.862 (95% CI: 0.843-0.880, 95% PI: 0.645-1.000) in the validation set. The average RQS across the included studies was 27.4%, revealing methodological limitations and an absence of standardization.

Conclusions:

Radiomics-based ML exhibited a higher predictive accuracy for the risk of recurrence in NSCLC. In contrast, combining radiomics with clinical features, ML failed to significantly improve the predictive accuracy.