Multiregional radiomics of brain metastasis can predict response to EGFR-TKI in metastatic NSCLC

FLAIR-based radiomics signature from brain-tumor interface for early prediction of response to EGFR-TKI therapy in NSCLC patients with brain metastasis

Objectives

Evaluating response to epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors (TKIs) is crucial in non-small cell lung cancer (NSCLC) patients with brain metastases (BM). To explore values of multi-sequence MRI in early assessing response to EGFR-TKIs in non-small cell lung cancer (NSCLC) patients with BM.

Approach

A primary cohort of 133 patients (January 2018 to March 2024) from center one and an external cohort of 52 patients (May 2017 to December 2022) from center two were established. Radiomics features were extracted from 4 mm brain-tumor interface (BTI) and whole BM region across T1-weighted contrast enhanced (T1CE) and T2-weighted (T2W) and T2 fluid-attenuated inversion recovery (T2-FLAIR) MRI sequences. The most relevant features were selected using the U test and least absolute shrinkage and selection operator (LASSO) method to develop the multi-sequence models based on BTI (RS-BTI-COM) and BM (RS-BM-COM). By integrating RS-BTI-COM with peritumoral edema volume (VPE), the combined model was built using logistic regression. Model performance was evaluated using the area under the ROC curve (AUC), sensitivity (SEN), specificity (SPE) and accuracy (ACC).

Main Results

The constructed RS-BTI-COM demonstrated a higher association with early response to EGFR-TKI therapy than RS-BM-COM. The combined RS-BTIplusVPE, incorporating BTI-based radiomics features and VPE, exhibited the highest AUCs (0.843-0.938), SPE (0.808-0.905) and ACC (0.712-0.875) in the training, internal validation, and external validation cohort, respectively.

Significance

The study developed a validated non-invasive model (RS-BTIplusVPE) based on integrating BTI-based radiomics features and VPE, which showed improved prediction of EGFR-TKI response in NSCLC patients with BM compared to tumor-focused models.

Machine learning in prediction of epidermal growth factor receptor status in non-small cell lung cancer brain metastases: a systematic review and meta-analysis

BACKGROUND

Epidermal growth factor receptor (EGFR) mutations are present in 10-60% of all non-small cell lung cancer (NSCLC) patients and are associated with dismal prognosis. Lung cancer brain metastases (LCBM) are a common complication of lung cancer. Predictions of EGFR can help physicians in decision-making and, through optimizing treatment strategies, can result in more favorable outcomes. This systematic review and meta-analysis evaluated the predictive performance of machine learning (ML)-based models in EGFR status in NSCLC patients with brain metastasis.

METHODS

On December 20, 2024, the four electronic databases, Pubmed, Embase, Scopus, and Web of Science, were systematically searched. Studies that evaluated EGFR status in patients with brain metastasis from NSCLC were included.

RESULTS

Twenty studies with 3517 patients with 6205 NSCLC brain metastatic lesions were included. The majority of the best-performance models were ML-based (70%, 7/10), and deep learning (DL)-based models comprised 30% (6/20) of models. The area under the curve (AUC) and accuracy (ACC) of the best-performance models ranged from 0.765 to 1 and 0.69 to 0.93, respectively. The meta-analysis of the best-performance model revealed a pooled AUC of 0.91 (95%CI: 0.88-0.93) and ACC of 0.82 (95%CI: 0.79-0.86) along with a pooled sensitivity of 0.87 (95%CI: 0.83-0.9), specificity of 0.86 (95%CI: 0.79-0.9), and diagnostic odds ratio (DOR) of 35.2 (95%CI: 21.2-58.4). The subgroup analysis did not show significant differences between ML and DL models.

CONCLUSION

ML-based models demonstrated promising predictive outcomes in predicting EGFR status. Applying ML-based models in daily clinical practice can optimize treatment strategies and enhance clinical and radiological outcomes.

The clinical implications and interpretability of computational medical imaging (radiomics) in brain tumors

Radiomics has widespread applications in the field of brain tumor research. However, radiomic analyses often function as a 'black box' due to their use of complex algorithms, which hinders the translation of brain tumor radiomics into clinical applications. In this review, we will elaborate extensively on the application of radiomics in brain tumors. Additionally, we will address the interpretability of handcrafted-feature radiomics and deep learning-based radiomics by integrating biological domain knowledge of brain tumors with interpretability methods. Furthermore, we will discuss the current challenges and prospects concerning the interpretability of brain tumor radiomics. Enhancing the interpretability of radiomics may make it more understandable for physicians, ultimately facilitating its translation into clinical practice. CRITICAL RELEVANCE STATEMENT: The interpretability of brain tumor radiomics empowers neuro-oncologists to make well-informed decisions from radiomic models. KEY POINTS: Radiomics makes a significant impact on the management of brain tumors in several key clinical areas. Transparent models, habitat analysis, and feature attribute explanations can enhance the interpretability of traditional handcrafted-feature radiomics in brain tumors. Various interpretability methods have been applied to explain deep learning-based models; however, there is a lack of biological mechanisms underlying these models.

Metastatic Lung Adenocarcinomas: Development and Evaluation of Radiomic-Based Methods to Measure Baseline Intra-Patient Inter-Tumor Lesion Heterogeneity

Radiomics has traditionally focused on individual tumors, often neglecting the integration of metastatic disease, particularly in patients with non-small cell lung cancer. This study sought to examine intra-patient inter-tumor lesion heterogeneity indices using radiomics, exploring their relevance in metastatic lung adenocarcinoma. Consecutive adults newly diagnosed with metastatic lung adenocarcinoma underwent contrast-enhanced CT scans for lesion segmentation and radiomic feature extraction. Three methods were devised to measure distances between tumor lesion profiles within the same patient in radiomic space: centroid to lesion, lesion to lesion, and primitive to lesion, with subsequent calculation of mean, range, and standard deviation of these distances. Associations between HIs, disease control rate, objective response rate to first-line treatment, and overall survival were explored. The study included 167 patients (median age 62.3 years) between 2016 and 2019, divided randomly into experimental (N = 117,546 lesions) and validation (N = 50,232 tumor lesions) cohorts. Patients without disease control/objective response and with poorer survival consistently systematically exhibited values of all heterogeneity indices. Multivariable analyses revealed that the range of primitive-to-lesion distances was associated with disease control in both cohorts and with objective response in the validation cohort. This metrics showed univariable associations with overall survival in the experimental. In conclusion, we proposed original methods to estimate the intra-patient inter-tumor lesion heterogeneity using radiomics that demonstrated correlations with patient outcomes, shedding light on the clinical implications of inter-metastases heterogeneity. This underscores the potential of radiomics in understanding and potentially predicting treatment response and prognosis in metastatic lung adenocarcinoma patients.

Habitat-Based Radiomics for Predicting EGFR Mutations in Exon 19 and 21 From Brain Metastasis

RATIONALE AND OBJECTIVES

To explore and externally validate habitat-based radiomics for preoperative prediction of epidermal growth factor receptor (EGFR) mutations in exon 19 and 21 from MRI imaging of non-small cell lung cancer (NSCLC)-originated brain metastasis (BM).

METHODS

A total of 170, 62 and 61 patients from center 1, center 2 and center 3, respectively were included. All patients underwent contrast-enhanced T1-weighted (T1CE) and T2-weighted (T2W) MRI scans. Radiomics features were extracted from the tumor active (TA) and peritumoral edema (PE) regions in each MRI slice. The most important features were selected by the least absolute shrinkage and selection operator regression to develop radiomics signatures based on TA (RS-TA), PE (RS-PE) and their combination (RS-Com). Receiver operating characteristic (ROC) curve analysis was performed to access performance of radiomics models for both internal and external validation cohorts.

RESULTS

10, four and six most predictive features were identified to be strongly associated with the EGFR mutation status, exon 19 and exon 21, respectively. The RSs derived from the PE region outperformed those from the TA region for predicting the EGFR mutation, exon 19 and exon 21. The RS-Coms generated the highest performance in the primary training (AUCs, RS-EGFR-Com vs. RS-exon 19-Com vs. RS-exon 21-Com, 0.955 vs. 0.946 vs. 0.928), internal validation (AUCs, RS-EGFR-Com vs. RS-exon 19-Com vs. RS-exon 21-Com, 0.879 vs. 0.819 vs. 0.882), external validation 1 (AUCs, RS-EGFR-Com vs. RS-exon 19-Com vs. RS-exon 21-Com, 0.830 vs. 0.825 vs. 0.822), and external validation 2 (AUCs, RS-EGFR-Com vs. RS-exon 19-Com vs. RS-exon 21-Com, 0.812 vs. 0.818 vs. 0.800) cohort.

CONCLUSION

The developed habitat-based radiomics model can be used to accurately predict the EGFR mutation subtypes, which may potentially guide personalized treatments for NSCLC patients with BM.

MRI radiomics-based interpretable model and nomogram for preoperative prediction of Ki-67 expression status in primary central nervous system lymphoma

Objectives

To investigate the value of interpretable machine learning model and nomogram based on clinical factors, MRI imaging features, and radiomic features to predict Ki-67 expression in primary central nervous system lymphomas (PCNSL).

Materials and methods

MRI images and clinical information of 92 PCNSL patients were retrospectively collected, which were divided into 53 cases in the training set and 39 cases in the external validation set according to different medical centers. A 3D brain tumor segmentation model was trained based on nnU-NetV2, and two prediction models, interpretable Random Forest (RF) incorporating the SHapley Additive exPlanations (SHAP) method and nomogram based on multivariate logistic regression, were proposed for the task of Ki-67 expression status prediction.

Results

The mean dice Similarity Coefficient (DSC) score of the 3D segmentation model on the validation set was 0.85. On the Ki-67 expression prediction task, the AUC of the interpretable RF model on the validation set was 0.84 (95% CI:0.81, 0.86; p < 0.001), which was a 3% improvement compared to the AUC of the nomogram. The Delong test showed that the z statistic for the difference between the two models was 1.901, corresponding to a p value of 0.057. In addition, SHAP analysis showed that the Rad-Score made a significant contribution to the model decision.

Conclusion

In this study, we developed a 3D brain tumor segmentation model and used an interpretable machine learning model and nomogram for preoperative prediction of Ki-67 expression status in PCNSL patients, which improved the prediction of this medical task.

Clinical relevance statement

Ki-67 represents the degree of active cell proliferation and is an important prognostic parameter associated with clinical outcomes. Non-invasive and accurate prediction of Ki-67 expression level preoperatively plays an important role in targeting treatment selection and patient stratification management for PCNSL thereby improving prognosis.

Distinguishing EGFR mutation molecular subtypes based on MRI radiomics features of lung adenocarcinoma brain metastases

OBJECTIVE

To explore the feasibility of identifying epidermal growth factor receptor (EGFR) mutation molecular subtypes in primary lesions based on the radiomics features of lung adenocarcinoma brain metastases using magnetic resonance imaging (MRI).

METHODS

We retrospectively analyzed clinical, imaging, and genetic testing data of patients with lung adenocarcinoma with EGFR gene mutations who had brain metastases. Three-dimensional radiomics features were extracted from contrast-enhanced T1-weighted images. The volume of interest was delineated and normalized using Z-score, dimensionality reduction was performed using principal component analysis, feature selection using Relief, and radiomics model construction using adaptive boosting as a classifier. Data were randomly divided into training and testing datasets at an 8:2 ratio. Five-fold cross-validation was conducted in the training set to select the optimal radiomics features and establish a predictive model for distinguishing between exon 19 deletion (19Del) and exon 21 L858R point mutation (21L858R), the two most common EGFR gene mutations. The testing set was used for external validation of the models. Model performance was evaluated using receiver operating characteristic curve and decision curve analyses.

RESULTS

Overall, 86 patients with 228 brain metastases were included. Patient age was identified as an independent predictor for distinguishing between 19Del and 21L858R. The area under the curve (AUC) values of the radiomics model in the training and testing datasets were 0.895 (95% confidence interval [CI]: 0.850-0.939) and 0.759 (95% CI: 0.0.614-0.903), respectively. The AUC for diagnosis of all cases using a combined model of age and radiomics was 0.888 (95% CI: 0.846-0.930), slightly higher than that of the radiomics model alone (0.866, 95% CI: 0.820-0.913), but without statistical significance (p=0.1626). In the decision curve analysis, both models demonstrated clinical net benefits.

CONCLUSIONS

The radiomics model based on MRI of lung adenocarcinoma brain metastases could distinguish between EGFR 19Del and 21L858R mutations in the primary lesion.

MRI radiomics for predicting intracranial progression in non-small-cell lung cancer patients with brain metastases treated with epidermal growth factor receptor tyrosine kinase inhibitors

AIM

To identify clinical and magnetic resonance imaging (MRI) radiomics predictors specialised for intracranial progression (IP) after first-line epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) treatment in non-small-cell lung cancer (NSCLC) patients with brain metastases (BMs).

MATERIALS AND METHODS

Seventy EGFR-mutated NSCLC patients with a total of 212 BMs who received first-line EGFR-TKI therapy were enrolled. Radiomics features were extracted from the BM regions on the pretreatment contrast-enhanced T1-weighted images, and the radiomics score (rad-score) of each BM was established based on the selected features. Furthermore, the mean rad-score derived from the average rad-score of all included BMs in each patient was calculated. Univariate and multivariate logistic regression analyses were performed to identify potential predictors of IP. Prediction models based on different predictors and their combinations were constructed, and nomogram based on the optimal prediction model was evaluated.

RESULTS

Thirty-three (47.1 %) patients developed IP, and the remaining 37 (52.9 %) patients were IP-free. EGFR-19del mutation (OR 0.19, 95 % CI 0.05-0.69), third-generation TKI treatment (OR 0.33, 95 % CI 0.16-0.67) and mean rad-score (OR 5.71, 95 % CI 1.65-19.68) were found to be independent predictive factors. Models based on these three predictors alone and in combination (combined model) achieved AUCs of 0.64, 0.64, 0.74, and 0.86 and 0.64, 0.64, 0.75, and 0.84 in the training and validation sets, respectively, and the combined model demonstrated optimal performance for predicting IP.

CONCLUSIONS

The model integrating EGFR-19del mutation, third-generation TKI treatment and mean rad-score had good predictive value for IP after EGFR-TKI treatment in NSCLC patients with BM.

Application of radiomics in diagnosis and treatment of lung cancer

Radiomics has become a research field that involves the process of converting standard nursing images into quantitative image data, which can be combined with other data sources and subsequently analyzed using traditional biostatistics or artificial intelligence (Al) methods. Due to the capture of biological and pathophysiological information by radiomics features, these quantitative radiomics features have been proven to provide fast and accurate non-invasive biomarkers for lung cancer risk prediction, diagnosis, prognosis, treatment response monitoring, and tumor biology. In this review, radiomics has been emphasized and discussed in lung cancer research, including advantages, challenges, and drawbacks.