Enhancement of Radiosurgical Treatment Outcome Prediction Using MRI Radiomics in Patients with Non-Small Cell Lung Cancer Brain Metastases

Abstract

The diagnosis of brain metastasis (BM) is commonly observed in non-small cell lung cancer (NSCLC) with poor outcomes. Accordingly, developing an approach to early predict BM response to Gamma Knife radiosurgery (GKRS) may benefit the patient treatment and monitoring. A total of 237 NSCLC patients with BMs (for survival prediction) and 256 patients with 976 BMs (for prediction of local tumor control) treated with GKRS were retrospectively analyzed. All the survival data were recorded without censoring, and the status of local tumor control was determined by comparing the last MRI follow-up in patients’ lives with the pre-GKRS MRI. Overall 1763 radiomic features were extracted from pre-radiosurgical magnetic resonance images. Three prediction models were constructed, using (1) clinical data, (2) radiomic features, and (3) clinical and radiomic features. Support vector machines with a 30% hold-out validation approach were constructed. For treatment outcome predictions, the models derived from both the clinical and radiomics data achieved the best results. For local tumor control, the combined model achieved an area under the curve (AUC) of 0.95, an accuracy of 90%, a sensitivity of 91%, and a specificity of 89%. For patient survival, the combined model achieved an AUC of 0.81, an accuracy of 77%, a sensitivity of 78%, and a specificity of 80%. The pre-radiosurgical radiomics data enhanced the performance of local tumor control and survival prediction models in NSCLC patients with BMs treated with GRKS. An outcome prediction model based on radiomics combined with clinical features may guide therapy in these patients.

The Impact of Resampling and Denoising Deep Learning Algorithms on Radiomics in Brain Metastases MRI

“…To do that, we used Bland-Altman plots and analysed the accuracy of the predicted values from radiomic model after DL processing in comparison to the original predicted values. Considering radiomic model from [29], only 2 and 3 predicted values are not accurate after resampling and denoising DL process, respectively, which represent accuracies of 95 and 92.5% (same results were obtained with radiomic model from [30]). These results highlight the impressive ability of DL to capture the shape and very precise features of the reference/high quality images during the training step and then re-inject them into new downsampled or denoised images.…”

“…To further explore the radiomic stability evaluation after DL processing, we evaluated the difference in predictive radiomic model results after DL resampling or fast images in comparison with reference images. Published overall survival [ 29 ] and classification [ 30 ] predictive models were used in this study. As shown in Figure 5 , Bland–Altman plots highlight important differences in predictive values obtained with reference and fast images (mean difference = −0.86, p < 0.001).…”

“…Finally, radiomic model validity after DL processing was evaluated as a final output. In this study we showed that whereas predictive values were strongly modified after fast acquisition, differences in predictive values after DL image processing in comparison to the reference was less important for one radiomic model [ 29 ] and with no difference for another one [ 30 ]. It is important to note that these radiomic models were obtained using a combination of radiomic features.…”

“…Concerning the radiomic models non-tested in this study, only 4/21 were unstable after DL resampling. More interestingly, the DL denoising model worked better with no unstable radiomic features for the two radiomic models [ 29 , 30 ] used in this study compared with only one for the radiomics model for classification used in the Qian study [ 51 ]. In regards to clinical implication, we assessed the impact of DL processing on the validity of predicted values from radiomic models.…”

“…To evaluate the impact of DL algorithms on radiomics, we compared predictive radiomic models values before and after DL algorithm processing. Two predictive models of radiomic were used base on [ 29 , 30 ], which are radiomic models for the prediction of treatment response (overall survival) of BM from NSCLC and BM classification. More details on the radiomic models can be found in Table S2 .…”