Length-scale study in deep learning prediction for non-small cell lung cancer brain metastasis

Objective: The MET receptor is mutated in 3-4% and amplified in 1-6% of patients with non-small cell lung cancer. The most common MET mutation is in exon 14, which results in deletion of the intracellular juxtamembrane domain of the receptor, leading to enhanced signaling. Mutation at Asp-1000 in the juxtamembrane region has also been reported and can lead to inhibition of apoptosis and cell proliferation. MET is involved in multiple pathways associated with radiation response. Therefore, we investigated the effects of inhibiting MET in combination with radiation in two preclinical non-small cell lung cancer (NSCLC) brain metastasis patient derived xenograft (PDX) models. Methods: Surgically obtained tissue was implanted subcutaneously into immunodeficient mice. Histology and DNA loci were compared between original tumor and PDX. DNA sequencing was performed on tumors for mutation analysis. In vivo growth responses to the MET inhibitor capmatinib or savolitinib with and without radiation were assessed. Radiation was delivered in 10 daily fractions of 2 Gy. Drug was administered by oral gavage daily 1 hour prior to radiation at a dose of 2.5 mg/kg for savaolitnib or 20 mg/kg for capmatinib for 10 days. Immunohistochemistry (IHC) was performed to evaluate MET signaling and proliferation. Results: PDXs were successfully established from two patients with MET mutated NSCLC brain metastases: a lung adenocarcinoma with a MET exon 14 skipping mutation and a lung sarcomatoid carcinoma with adenocarcinoma component with an Asp-1000 frameshift mutation. Morphologically, strong retention of cytoarchitectural features was observed between original patient tumors and PDXs. Short tandem repeat analysis confirmed 100% matching of alleles between patient tumors and PDXs. DNA sequencing confirmed METex14 and MET Asp-1000 frameshift mutation. In the METex14 PDX, savolitinib alone significantly inhibited tumor growth with a growth inhibition value of 46% (p<0.01). Combination of savolitinib and radiation significantly delayed growth compared to vehicle control, savolitinib alone or radiation alone (p<0.001). The absolute growth delay was 42.4 days for savolitinib plus radiation treatment compared to 9.3 days for savolitinib alone and 28.2 days for irradiation. In the MET Asp-1000 PDX, capmatinib and radiation significantly delayed growth compared the other treatment arms (p<0.01). IHC demonstrated inhibition of phospho-MET and pS6 and a decrease in Ki67 with MET inhibition. Conclusion: Inhibition of MET enhanced the effect of radiation in our two preclinical in vivo MET mutated NSCLC brain metastasis models. Additional studies are currently underway evaluating the mechanisms of radiation sensitization and the efficacy of this combination in MET amplified PDX models. Citation Format: Kwangok P. Nickel, Shrey Ramesh, Saahil Javeri, Nitin Somasundaram, Nan Sethakorn, Randall J. Kimple, Andrew M. Baschnagel. MET inhibition enhances the effect of radiation in MET mutated non-small cell lung cancer brain metastasis patient derived xenografts [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 207.

The development and characterization of patient derived xenografts from non-small cell lung cancer brain metastases

Targeted treatment options for non-small cell lung cancer (NSCLC) brain metastases (BMs) may be combined with stereotactic radiosurgery (SRS) to optimize survival. We assessed patient outcomes after SRS for NSCLC BMs, identifying survival trajectories associated with targetable mutations. In this retrospective time-dependent analysis, we analyzed median overall survival of patients who received ≥ 1 SRS courses for BM from NSCLC from 2001 to 2021. We compared survival of patients with and without targetable mutations based on clinical variables and treatment. Among the 213 patients included, 87 (40.8%) had targetable mutations-primarily EGFR (22.5%)-and 126 (59.2%) did not. Patients with targetable mutations were more often female (63.2%, p <.001) and nonsmokers (58.6%, p <.001); had higher initial lung-molGPA (2.0 vs. 1.5, p <.001) and lower cumulative tumor volume (3.7 vs. 10.6 cm3, p <.001); and received more concurrent (55.2% vs. 36.5%, p =.007) and total (median 3 vs. 2, p <.001) systemic therapies. These patients had lower mortality rates (74.7% vs. 91.3%, p <.001) and risk (HR 0.298 [95%CI 0.190-0.469], p <.001) and longer median overall survival (20.2 vs. 7.4 months, p <.001), including survival ≥ 3 years (p =.001). Survival was best predicted by SRS with tumor resection in patients with non-targetable mutations (HR 0.491 [95%CI 0.318-757], p =.001) and by systemic therapy with SRS for those with targetable mutations (HR 0.124 [95%CI 0.013-1.153], p =.067). The presence of targetable mutations enhances survival in patients receiving SRS for NSCLC BM, particularly when used with systemic therapies. Survival for patients without targetable mutations was longest with SRS and surgical resection. These results inform best practices for managing patients with NSCLC BM based on driver mutation status.

Multimodal radiomic analysis to determine high-Consistency prognostic phenotypes associated with epidermal growth factor receptor mutations in non-small cell lung cancer brain metastases.

Immune-Depleted Phenotype in Non-Small Cell Lung Cancer Brain Metastases Predicts for Worse Overall Survival

Stereotactic radiosurgery/radiotherapy (SRS/SRT) and novel systemic treatments, such as tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs), have demonstrated to be effective in managing brain metastases in non-small cell lung cancer (NSCLC). However, the optimal treatment sequence of SRS/SRT and TKI/ICI remains uncertain. This retrospective monocentric analysis addresses this question by comparing the outcomes of patients with NSCLC brain metastases who received upfront SRS/SRT versus those who were initially treated with TKI/ICI. All patients treated with SRS/SRT and TKI/ICI for NSCLC brain metastases were collected from a clinical database. The patients who received first-line TKI or ICI for the treatment of brain metastases were then selected for further analysis. Within this cohort, a comparative analysis between upfront SRS/SRT and patients initially treated with TKI/ICI was conducted, assessing key parameters such as overall survival (OS), intracranial progression-free survival (iPFS) and treatment-related toxicity. Both OS and iPFS were defined as the time from SRS/SRT to either death or disease progression, respectively. The analysis encompassed 54 patients, of which 34 (63.0%) patients received SRS/SRT and TKI/ICI as their first-line therapy. Of the latter, 17 (50.0%) patients received upfront SRS/SRT and 17 (50.0%) were initially treated with TKI/ICI; 24 (70.6%) received SRS/SRT and ICI, and 10 (29.4%) received SRS/SRT and TKI. The cohorts did not significantly differ in the univariable analyses for the following parameters: sex, age, histology, molecular genetics, disease stage at study treatment, performance status, number of brain metastases, treatment technique, tumor volume, target volume, disease progression, radiation necrosis, dosimetry. While no significant differences were found in terms of iPFS and OS between patients treated with upfront SRS/SRT and patients initially treated with TKI, upfront SRS/SRT demonstrated significantly superior OS when compared to patients initially treated with ICI (median OS not reached vs. 17.5 months; mean 37.8 vs. 23.6 months; P=0.03) with no difference in iPFS. No significant differences in treatment-related toxicity were observed among the cohorts. In this retrospective, single-center cohort study, patients treated with upfront SRS/SRT demonstrated significantly longer OS compared to patients initially treated with ICI in the cohort receiving first-line therapy for brain metastases. However, given the retrospective design and the limited cohort size, definitive conclusions cannot be drawn from these findings. Nevertheless, the results suggest that the timing of SRS/SRT may play an important role in treatment outcomes. Further investigation, preferably through prospective randomized trials, is warranted to provide more conclusive answers to this important question.

Several reports have described extended survival after aggressive surgical treatment for non-small cell lung cancer (NSCLC) and synchronous brain metastasis. This retrospective analysis assesses the prognostic factors in this population. We reviewed retrospectively the medical records of 29 patients with synchronous brain metastasis from NSCLC, who underwent surgical treatment in our institution between 1980 and 2008. All patients underwent chest surgery to remove the primary lesion. The impact of several variables on survival was assessed. The median follow-up period was 9.6 months and the 5-year survival rate from the time of lung cancer resection was 20.6 %. Univariate analysis demonstrated that the carcinoembryonic antigen (CEA) level, primary tumor size, and the presence of lymph node involvement were predictive of overall survival (p < 0.05). Multivariate analysis also identified those factors to be independent favorable prognostic factors. Although the survival of patients with brain metastasis from non-small cell lung cancer remains poor, surgical resection may benefit a select group of patients, particularly those with a normal CEA level, small tumor size, and node-negative status.

Predicting clinical and radiological outcomes of epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs) in patients with non-small cell lung cancer (NSCLC) and brain metastases (BMs) is crucial for effective patient management. Machine learning (ML)-based models have increasingly been utilized to predict EGFR-TKI response in patients with lung cancer brain metastasis (LCBM). In this study, we aimed to evaluate the predictive performance of ML-based models for EGFR-TKI response prediction. A comprehensive literature search was conducted using PubMed, Embase, Scopus, and Web of Science from database inception to April 25, 2025. Studies that developed ML-based models to predict EGFR-TKI response were included. Eight studies involving 1322 LCBM patients were included. The included studies used logistic regression (LR), LR with least absolute shrinkage and selection operator (LASSO), decision tree (DT), and a Cox-based deep learning model (DL-Cox). The meta-analysis revealed a pooled area under the curve (AUC) of 0.84 (95% CI 0.78-0.91) and accuracy (ACC) of 0.75 (95% CI 0.62-0.88) with a sensitivity (SEN) of 0.82 (95% CI 0.77-0.87) and a specificity (SPE) of 0.73 (95% CI 0.66-0.80) for prediction of EGFR-TKI response. The meta-analysis of diagnostic odds ratios (DOR) exhibited a pooled DOR of 12.41 (95% CI 7.32-21.04). ML-based models show promising ability to predict EGFR-TKI response in LCBM, supporting their potential to guide treatment selection. However, their use in clinical practice remains limited by small retrospective datasets and lack of external validation.

Stereotactic radiotherapy (SRT) of brain metastases (BM) results are often reported in the heterogeneous primitive population. Here, we report our experience in consecutively treated patients who underwent SRT alone for BM from non-small cell lung cancer (NSCLC). This retrospective analysis included consecutive patients with no history of cerebral treatment who underwent Cyberknife™ SRT for BM from NSCLC in our institution from 2007 to 2016. One hundred patients were included in the study, with a median follow-up of 33 months (20–64). Mean age was 63 years (SD ± 10); 88% had Karnofsky Performance Status (KPS) > 70; 67% had unique BM; 18 patients received single-fraction SRT (20–25 Gy), and 82 received hypo-fractionated SRT (HSRT) (24–36 Gy in 3–5 fractions). We reported a complication rate of 17% (2% of G3-4). Median survival was 10.1 months [95% confidence interval (CI) 7.8–13.9]. At 1 year, local and cerebral control rates were respectively 78.7% (95% CI 70–86.5%) and 43% (95% CI 33.5–53%). Thirty patients underwent salvage treatment (whole brain radiation therapy, n = 13; SRT, n = 14; surgery, n = 3). Cyberknife™-based SRT is an effective treatment associated with high local control rate with low morbidity for patients with NSCLC’s BM. Close follow-up is necessary to perform salvage treatment.

Multisequence MRI-based radiomics nomogram for early prediction of osimertinib resistance in patients with non-small cell lung cancer brain metastases

Brain metastasis is a serious complication of non-small cell lung cancer (NSCLC) affecting up to 40% of NSCLC patients. A subset of NSCLC tumors has mutations in the epidermal growth factor receptor (EGFR) gene, and determination of tumor EGFR mutation status is essential in guiding treatment decisions, as it directly affects the treatment approach. Patients with EGFR-mutated NSCLC have a higher cumulative incidence of brain metastases, and are especially sensitive to EGFR tyrosine kinase inhibitors (TKIs). Patients with newly diagnosed EGFR-mutated lung cancer presenting to a neurosurgeon with a new diagnosis of brain metastases now have a variety of treatment options available, including whole brain radiation therapy, stereotactic radiosurgery, surgical resection, chemotherapy, and targeted therapeutics such as the EGFR TKIs. In this review, we discuss the impact of EGFR mutation status on brain and leptomeningeal metastasis treatment considerations. Additionally, we present clinical cases of patients treated with EGFR TKIs alone and in combination with other therapies to highlight treatment alternatives.

The prognosis of patients suffering from non-small cell lung carcinomas (NSCLC) worsens significantly when brain metastasis occurs. Seeding to the brain usually happens relatively early in the course of disease and therefore, new therapies anticipating this complication would result in considerable improvement in outcomes. In this review, we address recent molecular data of NSCLC with a focus on the risk of the formation of brain metastasis. Included is new data on the involvement of miRNAs and lncRNAs in the rise of the cerebral seeding of NSCLC. We summarize novel therapeutic approaches developed in the light of these recent molecular discoveries.

Small cell lung cancer (SCLC) brain metastases (BM) are common at presentation and relapse. Historically, BM were managed with whole brain radiotherapy (WBRT). However, WBRT is associated with significant toxicity, and so interest in the focal management of BM (focal radiotherapy including stereotactic radiosurgery (SRS) and surgery) is increasing. A systematic literature search of PubMed, Embase, and CENTRAL databases and clinicaltrials.gov was conducted as per the Preferred Reporting Items for Systematic Review and Meta-Analyses guidelines to obtain all published articles, abstracts, and ongoing studies regarding focal management of SCLC BM. We identified 59 eligible articles, 6 related to surgical BM treatment (4 case studies, 2 retrospective case studies), and 53 regarding focal radiotherapy (4 case reports, 9 retrospective cohort studies, 32 retrospective case series, 2 prospective cohort trials, 6 ongoing trials). Several trials (n = 6) are ongoing investigating outcomes following focal radiotherapy for SCLC BM, alone and versus WBRT. Whilst limited by their retrospective design, and lack of information regarding local control, first line focal radiotherapy ± WBRT seems to be associated with better survival than WBRT alone. Additionally, salvage focal radiotherapy can achieve 14.3 months median overall survival, outperforming the 3 months median overall survival with salvage WBRT. In highly selected patients, neurosurgical resection of brain metastases may play a role in optimizing local control and/or relieving BM symptoms. Focal treatments for SCLC BM are being used more frequently. However, randomized trial evidence for their safety and effectiveness are still pending.

Automated image classification is an essential task of the computer vision field. The tagging of images into a set of predefined groups is referred to as image classification. The implementation of computer vision to automate image classification would be beneficial because manual image evaluation and identification can be time-consuming, particularly when there are many images of different classes. Deep learning approaches are proven to overperform existing machine learning techniques in many fields in recent years, and computer vision is one of the most notable examples. The very deep neural network (VDCNN) is a powerful deep learning model for image classification, and this paper examines it briefly using MNIST handwritten digit dataset. This dataset is used to prove the efficacy of very deep neural networks over other deep learning models. The proposed study aims to comprehend the very deep neural network architecture used to accomplish a handwritten digit recognition task. The feasibility of the proposed model is evaluated using mean accuracy, validation accuracy, and standard deviation. The study results of the very deep neural network model are compared to a convolutional neural network and convolutional neural network with batch normalization. According to the results of the comparison study, very deep neural networks achieve high accuracy of 99.1% for a handwritten dataset. The outcome of the proposed work is used to interpret how well a very deep neural network performs when compared to the other two models of deep neural networks. This proposed architecture may be used to automate the classification of handwritten digits dataset.

Survey on Deep Neural Networks in Speech and Vision Systems