Predictive Value of Radiomics Features Extracted from preoperative CT Images for Assessing Response to Thermal Ablation in Patients with Malignant Pulmonary Lesions: A Systematic Review, Quality Assessment, and Meta-Analysis.

To determine the prognostic value of Radiomics features in determining objective treatment response in patients with pulmonary masses undergoing thermal ablation.

Materials and Methods:

A systematic search of literature was performed on relevant databases, and all studies underwent screening by two readers. The Quadas-2 and METhodological RadiomICs Score (METRICS) tools were used to assess the quality and clinical applicability of the presented modules. A random effects meta-analysis was performed on pooled data. Clinical applicability of the modules were determined based on Fagan’s nomogram.

Results:

A total of 13 studies including 622 patients were included in the pooled analysis. M-RECIST was the most common criteria used for objective response (11/13). Six-months follow up was used to determine if the lesions had responded to thermal ablation. The mean METRICS score of the included studies was 24 ± 13.2, showing an overall low Radiomics methodologic quality. The included studies had low risk of bias based on the QUADAS-2 tool. Major contributors to low methodologic quality consisted to lack of multi-dataset validation, reliance on manual segmentation and lack of the establishment of radiomics-biological correlates.

The pooled sensitivity and specificity in predicting objective response was 89.1 (87-91) and 86.4 (83-89), respectively. The area under the SROC curve was 0.86 (84-90). The Fagan nomogram showed a skewed sample size towards responders, with 65% of the patient population showing objective response. Implementation of a radiomics module was able to decrease a negative likelihood ratio of 31% to 15%.

Conclusion:

In conclusion, this study highlights the potential prognostic value of radiomics features in predicting treatment response in patients with pulmonary masses undergoing thermal ablation. Despite the overall low methodological quality of radiomics approaches, the pooled analysis demonstrates high sensitivity and specificity. The findings suggest that integrating radiomics could significantly improve clinical decision-making, although further validation and methodological refinement are necessary to enhance reliability and applicability in clinical settings.