The evolutionary processes that underlie the marked sensitivity of small cell lung cancer (SCLC) to chemotherapy and rapid relapse are unknown^1–3. Here we determined tumour phylogenies at diagnosis and throughout chemotherapy and immunotherapy by multiregion sequencing of 160 tumours from 65 patients. Treatment-naive SCLC exhibited clonal homogeneity at distinct tumour sites, whereas first-line platinum-based chemotherapy led to a burst in genomic intratumour heterogeneity and spatial clonal diversity. We observed branched evolution and a shift to ancestral clones underlying tumour relapse. Effective radio- or immunotherapy induced a re-expansion of founder clones with acquired genomic damage from first-line chemotherapy. Whereas TP53 and RB1 alterations were exclusively part of the common ancestor, MYC family amplifications were frequently not constituents of the founder clone. At relapse, emerging subclonal mutations affected key genes associated with SCLC biology, and tumours harbouring clonal CREBBP/EP300 alterations underwent genome duplications. Gene-damaging TP53 alterations and co-alterations of TP53 missense mutations with TP73, CREBBP/EP300 or FMN2 were significantly associated with shorter disease relapse following chemotherapy. In summary, we uncover key processes of the genomic evolution of SCLC under therapy, identify the common ancestor as the source of clonal diversity at relapse and show central genomic patterns associated with sensitivity and resistance to chemotherapy.

Background: Brain metastases (BM) frequently cause focal neurological deficits leading to a reduced Karnofsky performance score (KPS). Since KPS is routinely used to guide the choice of adjuvant therapy, we hypothesized that improving KPS by surgical resection may improve the chance for adjuvant treatment and ultimately result in better survival. We therefore analyzed the course of a large cohort undergoing resection of symptomatic brain metastases in the context of further treatment and clinical outcome.Patients and methods: In a bi‐centric retrospective analysis we retrieved baseline, clinical, and treatment‐related parameters of patients operated on BM between 2010 and 2019. Survival was calculated using Kaplan‐Meier estimates; prognostic factors for survival were analyzed by Log‐rank test and Cox proportional hazards.Results: We included 750 patients with a median age of 61 (19‐87) years. The functional status was significantly improved by surgical resection, with a median preoperative (KPS) of 80 (10‐100) increasing to 90 (0‐100) after surgery (P < .0001). Moreover, surgery improved the RTOG recursive partitioning analysis (RPA) class from III to I/II in 82 patients. Postoperative local radiotherapy and systemic treatment were associated with significantly longer survival (P < .0001 for each). Systemic treatment was provided significantly more frequently in patients with a fair postoperative clinical status (KPS ≥ 70; P < .0001). The postoperative clinical status, postoperative radiotherapy, systemic treatment, controlled systemic disease and < 4 BM were independent predictors for survival.Conclusion: The resection of symptomatic BM may restore clinical status, so enhancing the likelihood of receiving adjuvant treatment, and therefore leading to improved overall survival.

Background: Non‐small cell lung cancer (NSCLC) is the most common tumor entity spreading to the brain and up to 50% of patients develop brain metastases (BMs). Detection of BMs on MRI is challenging with an inherent risk of missed diagnosis.Purpose: To train and evaluate a deep learning model (DLM) for fully automated detection and 3D segmentation of BMs in NSCLC on clinical routine MRI.Study Type: Retrospective.Population: Ninety‐eight NSCLC patients with 315 BMs on pretreatment MRI, divided into training (66 patients, 248 BMs) and independent test (17 patients, 67 BMs) and control (15 patients, 0 BMs) cohorts.Field Strength/Sequence: T₁‐/T₂‐weighted, T₁‐weighted contrast‐enhanced (T₁CE; gradient‐echo and spin‐echo sequences), and FLAIR at 1.0, 1.5, and 3.0 T from various vendors and study centers.Assessment: A 3D convolutional neural network (DeepMedic) was trained on the training cohort using 5‐fold cross‐validation and evaluated on the independent test and control sets. Three‐dimensional voxel‐wise manual segmentations of BMs by a neurosurgeon and a radiologist on T₁CE served as the reference standard.Statistical Tests: Sensitivity (recall) and false positive (FP) findings per scan, dice similarity coefficient (DSC) to compare the spatial overlap between manual and automated segmentations, Pearson's correlation coefficient (r) to evaluate the relationship between quantitative volumetric measurements of segmentations, and Wilcoxon rank‐sum test to compare the volumes of BMs. A P value <0.05 was considered statistically significant.Results: In the test set, the DLM detected 57 of the 67 BMs (mean volume: 0.99 ± 4.24 cm³), resulting in a sensitivity of 85.1%, while FP findings of 1.5 per scan were observed. Missed BMs had a significantly smaller volume (0.05 ± 0.04 cm³) than detected BMs (0.96 ± 2.4 cm³). Compared with the reference standard, automated segmentations achieved a median DSC of 0.72 and a good volumetric correlation (r = 0.95). In the control set, 1.8 FPs/scan were observed.Data Conclusion: Deep learning provided a high detection sensitivity and good segmentation performance for BMs in NSCLC on heterogeneous scanner data while yielding a low number of FP findings.Level of Evidence: 3Technical Efficacy Stage: 2

To evaluate the effect of timing of radiotherapy (RT) on survival in patients with newly diagnosed primary glioblastoma (GBM) treated with the same therapeutical protocol. Patients with newly diagnosed primary GBM treated with the same therapeutical scheme between 2010 and 2015 in our institution were retrospectively reviewed. The population was trichotomized based on the time interval from surgery till initiation of RT (< 28 days, 28–33 days, > 33 days). Kaplan-Meier and Cox regression analyses were used to compare progression free survival (PFS) and overall survival (OS) between the groups. The influence of various extensively studied prognostic factors on survival was assessed by multivariate analysis. One-hundred-fifty-one patients met the inclusion criteria. Between the three groups no significant difference in PFS (p = 0.516) or OS (p = 0.902) could be demonstrated. Residual tumor volume (RTV) and midline structures involvement were identified as independent prognostic factors of PFS while age, O-6-Methylguanine Methyltransferase (MGMT) status, Ki67 index, RTV and midline structures involvement represented independent predictors of OS. Patients starting RT after a prolonged delay (> 48 days) exhibited a significantly shorter OS (p = 0.034). Initiation of RT within a timeframe of 48 days is not associated with worsened survival. A prolonged delay (> 48 days) may be associated with worse OS. RT should neither be delayed, nor forced, but should rather start timely, as soon as the patient has recovered from surgery.

BACKGROUND AND PURPOSE: Malignant melanoma is an aggressive skin cancer in which brain metastases are common. Our aim was to establish and evaluate a deep learning model for fully automated detection and segmentation of brain metastases in patients with malignant melanoma using clinical routine MR imaging. MATERIALS AND METHODS: Sixty-nine patients with melanoma with a total of 135 brain metastases at initial diagnosis and available multiparametric MR imaging datasets (T1-/T2-weighted, T1-weighted gadolinium contrast-enhanced, FLAIR) were included. A previously established deep learning model architecture (3D convolutional neural network; DeepMedic) simultaneously operating on the aforementioned MR images was trained on a cohort of 55 patients with 103 metastases using 5-fold cross-validation. The efficacy of the deep learning model was evaluated using an independent test set consisting of 14 patients with 32 metastases. Manual segmentations of metastases in a voxelwise manner (T1-weighted gadolinium contrast-enhanced imaging) performed by 2 radiologists in consensus served as the ground truth. RESULTS: After training, the deep learning model detected 28 of 32 brain metastases (mean volume, 1.0 [SD, 2.4] cm³) in the test cohort correctly (sensitivity of 88%), while false-positive findings of 0.71 per scan were observed. Compared with the ground truth, automated segmentations achieved a median Dice similarity coefficient of 0.75. CONCLUSIONS: Deep learning–based automated detection and segmentation of brain metastases in malignant melanoma yields high detection and segmentation accuracy with false-positive findings of <1 per scan.