An accessible deep learning tool for voxel-wise classification of brain malignancies from perfusion MRI.

Noninvasive differential diagnosis of brain tumors is currently based on the assessment of magnetic resonance imaging (MRI) coupled with dynamic susceptibility contrast (DSC). However, a definitive diagnosis often requires neurosurgical interventions that compromise patients' quality of life. We apply deep learning on DSC images from histology-confirmed patients with glioblastoma, metastasis, or lymphoma. The convolutional neural network trained on ∼50,000 voxels from 40 patients provides intratumor probability maps that yield clinical-grade diagnosis. Performance is tested in 400 additional cases and an external validation cohort of 128 patients. The tool reaches a three-way accuracy of 0.78, superior to the conventional MRI metrics cerebral blood volume (0.55) and percentage of signal recovery (0.59), showing high value as a support diagnostic tool. Our open-access software, Diagnosis In Susceptibility Contrast Enhancing Regions for Neuro-oncology (DISCERN), demonstrates its potential in aiding medical decisions for brain tumor diagnosis using standard-of-care MRI.

Dose-Dependent Atrophy in Bilateral Amygdalae and Nuclei After Brain Radiation Therapy and Its Association With Mood and Memory Outcomes on a Longitudinal Clinical Trial.

PURPOSE: Amygdalae are bilateral, almond-shaped structures located anterior to the hippocampi, critical to limbic system functions of emotional processing and memory consolidation. The amygdalae are heterogeneous, composed of multiple nuclei with distinct structural and functional properties. We prospectively assessed associations between longitudinal changes in amygdala morphometry, including component nuclei, and functional outcomes in patients with primary brain tumors receiving radiation therapy (RT). METHODS AND MATERIALS: On a prospective longitudinal trial, 63 patients underwent high-resolution volumetric brain magnetic resonance imaging and testing for mood (Beck Depression Inventory and Beck Anxiety Inventory), memory (Brief Visuospatial Memory Test-Revised [BVMT] Total Recall and Delayed Recall; Hopkins Verbal Learning Test-Revised [HVLT] Total Recall and Delayed Recall), and health-related quality-of-life outcomes (Functional Assessment of Cancer Therapy-Brain Social/Family Well-Being and Emotional Well-Being) at baseline and 3, 6, and 12 months after RT. Amygdalae, including 8 nuclei, were autosegmented bilaterally using validated techniques. Linear mixed-effects models assessed longitudinal change in amygdalae and nuclei volumes and associations with dose and outcomes. Wilcoxon rank sum tests compared amygdala volume change between patient groups with worse and more stable outcomes at each time point. RESULTS: Atrophy was found in the right amygdala at 6 months (P = .001) and the left amygdala at 12 months (P = .046). A higher dose was associated with atrophy of the left amygdala (P = .013) at 12 months. The right amygdala showed dose-dependent atrophy at 6 months (P = .016) and 12 months (P = .001). Worse BVMT-Total, HVLT-Total, and HVLT-Delayed performance was associated with smaller left lateral (P = .014, P = .004, and P = .007, respectively) and left basal (P = .034, P = .016, and P = .026, respectively) nuclei volumes. Increased anxiety at 6 months was associated with greater combined (P = .031) and right (P = .007) amygdala atrophy. Greater left amygdala atrophy (P = .038) was noted in patients with decreased emotional well-being at 12 months. CONCLUSIONS: Bilateral amygdalae and nuclei undergo time- and dose-dependent atrophy after brain RT. Atrophy in amygdalae and specific nuclei was associated with poorer memory, mood, and emotional well-being. Amygdalae-sparing treatment planning may preserve neurocognitive and neuropsychiatric outcomes in this population.

EGFR mutant locally advanced non-small cell lung cancer is at increased risk of brain metastasis.

BACKGROUND AND PURPOSE: Small studies of primarily metastatic non-small cell lung cancer (NSCLC) have suggested an association between EGFR mutation (EGFR+) and likelihood of brain metastasis. However, these studies are confounded by follow-up time bias. We performed a competing risk analysis of brain metastasis in a more uniform locally advanced NSCLC (LA-NSCLC) cohort with known tumor genotype. MATERIALS AND METHODS: Between 2002 and 2014, 255 patients with LA-NSCLC underwent tumor genotyping for EGFR, ALK and/or KRAS (180 patients had follow-up brain imaging). Cumulative incidence and Fine-Gray regression were performed on clinical variables including genotype and risk of brain metastasis, with death as a competing event. RESULTS: The proportion of tumors with aberrations in EGFR, ALK and KRAS were 17%, 4% and 28%, respectively. The median follow-up was 68 months. On multivariate analysis, EGFR+ was significantly associated with risk of brain metastasis in the full patient cohort (HR 2.04, 95% CI 1.22-3.39, p = 0.006) as well as in the subset of patients with brain follow-up imaging (HR 1.91. 95% CI 1.17-3.13, p = 0.01). This translated to a higher cumulative incidence of brain metastasis in EGFR+ patients at 3 and 5 years (33.3% vs. 23.2 and 43.8% vs. 24.2%, p = 0.006). CONCLUSION: Patients with EGFR+ LA-NSCLC have a significantly higher likelihood of developing brain metastasis after standard combined modality therapy, independent of their longer overall survival. This high-risk genotypic subgroup may benefit from routine surveillance with brain MRI to allow early salvage with targeted systemic- and/or radiation-therapies.

Clinical Outcomes After Lung Stereotactic Body Radiation Therapy in Patients With or Without a Prior Lung Resection.

OBJECTIVES: Tumor control (TC), toxicity and survival, following stereotactic body radiation therapy (SBRT) were compared between patients with and without a prior lung resection (PLR). MATERIALS AND METHODS: The study is comprised of 130 patients with 141 peripheral tumors treated with SBRT at our institution from 2009 to 2013. Primary TC and lobar control (LC) were defined per RTOG 0236. Toxicity was scored using Common Terminology Criteria for Adverse Events version 4.0. Survival/TC and toxicity were compared between patients with and without PLR using the Kaplan-Meier method and cumulative incidence, respectively. Fine and Gray regression was used for univariable/multivariable analysis for radiation pneumonitis (RP). RESULTS: Of the 130 patients with median age 70 years (range, 42 to 93 y), 50 had undergone PLR (median time between PLR and SBRT: 33 mo; range, 1 to 206), including pneumonectomy (12%), lobectomy (46%), wedge resection (42%). With a median follow-up of 21 months in survivors, the PLR group had better TC (1-y 100% vs. 93%; P<0.01) and increased grade ≥2 (RP; 1-y 12% vs. 1%; P<0.01). OS was not significantly different between the 2 groups (1-y 91% vs. 85%; P=0.24). On univariable/multivariable analyses, biologically effective dose was associated with TC (hazard ratios, 0.97; 95% confidence interval, 0.94-0.999; P=0.04). Chemotherapy use was associated with grade ≥2 RP for all patients (hazard ratios, 14.92; 95% confidence interval, 5.68-39.21; P<0.0001) in multivariable analysis. PLR was not associated with increased RP in multivariable analysis. CONCLUSIONS: Patients with PLR who receive lung SBRT for lung tumors have high local control and relatively low toxicity. SBRT is an excellent option to treat second lung tumors or pulmonary metastases in patients with PLR.

Outcomes by EGFR, KRAS, and ALK Genotype After Combined Modality Therapy for Locally Advanced Non-Small-Cell Lung Cancer.

PURPOSE: In 699 patients with locally advanced non-small-cell lung cancer (NSCLC) treated with radiation therapy as part of combined modality therapy, we compared outcomes among genotyped and ungenotyped patients and by tumor genotype status (EGFR, KRAS, and ALK). PATIENTS AND METHODS: Genotyping was performed in 250 patients: EGFR+ (19%), KRAS+ (32%), ALK+ (9%), and wild type (WT-/-/-; 40%). Outcomes were analyzed using the Kaplan-Meier method and Cox regression. RESULTS: With a median follow-up of 48.2 months among genotyped patients, median overall survival (OS) was significantly longer for EGFR+ and ALK+ compared with KRAS+ and WT-/-/- (55.8 months v not reached v 28.0 v 33.2 months; P = .02). There was no difference in progression-free survival (median, 15.3 v 13.7 v 13.0 v 14.5 months; P = .47) or in freedom from distant metastases by genotype (3-year estimates: 42% v 49% v 27% v 25%; P = .25). There was higher freedom from locoregional recurrence (LRR) for EGFR+ tumors and lower freedom from LRR in ALK+ tumors, compared with KRAS+ and WT-/-/- tumors (3-year: 77% v 38% v 49% v 46%). In multivariable analysis, ALK+ remained associated with increased OS (HR, 0.32; 95% CI, 0.12 to 0.87; P = .03), and EGFR+ was associated with decreased LRR (HR, 0.47; 95% CI, 0.24 to 0.92; P = .03). Analysis of post-recurrence survival demonstrated that EGFR+/ALK+ patients treated with appropriate tyrosine kinase inhibitors had higher OS compared with other groups. CONCLUSION: In this series of locally advanced NSCLC treated with combined modality therapy, EGFR+ and ALK+ were associated with higher OS, whereas LRR was lower in EGFR+ patients, and the risk of distant metastases was high in all subgroups. The outcomes and patterns of failure in genotypic subgroups of NSCLC from this study can inform the design of future trials integrating targeted therapies.

Transcriptome evaluation of the relation between body mass index and prostate cancer outcomes.

BACKGROUND: Large epidemiological studies indicate that an increased body mass index (BMI) is associated with increased prostate cancer (PCa) mortality. Data indicate that there is no association between elevated metabolic pathway proteins and PCa mortality. There are no published studies evaluating the relation between BMI and metabolic pathways with respect to PCa outcomes with a genomics approach. METHODS: The Decipher Genomic Resource Information Database was queried for patients who had undergone prostatectomy and had BMI information available. These patients came from Thomas Jefferson University (TJU) and Johns Hopkins Medical Institution (JHMI); the latter provided 2 cohorts (I and II). A high-BMI group (≥30 kg/m2 ) and a low-BMI group (<25 kg/m2 ) were identified, and genomic data were interrogated for differentially expressed genes with an interquartile range filter and a Wilcoxon test. P values were adjusted for multiple testing with the Benjamini-Hochberg false-discovery rate method. RESULTS: A total of 477 patients with a median follow-up of 108 months had BMI information available. Two genes were found to interact with BMI in both the JHMI I cohort and the TJU cohort, but there was no statistical significance after adjustments for multiple comparisons. Aberrant metabolic gene expression was significantly correlated with distant metastases (P < .05). No relation was found between BMI and metastases or overall survival (both P values > .05). CONCLUSIONS: In a genomic analysis of prostatectomy specimens, metabolic gene expression, but not BMI, was associated with PCa metastases. Cancer 2017;123:2240-2247. © 2017 American Cancer Society.

Advanced nodal stage predicts venous thromboembolism in patients with locally advanced non-small cell lung cancer.

OBJECTIVES: Patients with non-small cell lung cancer (NSCLC) are known to be at high risk for venous thromboembolism (VTE), but previous studies have not specifically analyzed locally advanced disease. We performed a retrospective VTE risk analysis in a cohort of locally advanced NSCLC treated with definitive intent including radiation therapy. MATERIALS AND METHODS: The cohort consisted of 629 patients with stage II-III NSCLC treated at a single institution from January 2003 to December 2012. All patients received treatment with curative intent, including radiation therapy. Fine and Gray's competing-risks regression model, accounting for death and distant metastasis as competing risks, was used to identify significant predictors of VTE risk, and cumulative incidence estimates were generated using the competing-risks model. RESULTS AND CONCLUSION: At a median follow-up of 31 months, 127 patients developed a VTE, with 80% of events occurring in the first year after treatment initiation. 1-year and 3-year overall cumulative incidence estimates were 13.5% and 15.4%, respectively. On univariate analysis, stage IIIB and N3 nodal disease were associated with increased VTE risk. In the final multivariable model, N3 nodal disease was associated with increased VTE risk (Hazard ratio 1.64; 95% CI 1.06-2.54; p=0.027). In conclusion, patients with locally advanced NSCLC are at high risk for VTE, especially in the first year after treatment initiation, with a 1-year cumulative incidence of 13.5%. N3 nodal staging was associated with significantly higher VTE risk compared to N0-N2 staging.

CT-based radiomic signature predicts distant metastasis in lung adenocarcinoma.

BACKGROUND AND PURPOSE: Radiomics provides opportunities to quantify the tumor phenotype non-invasively by applying a large number of quantitative imaging features. This study evaluates computed-tomography (CT) radiomic features for their capability to predict distant metastasis (DM) for lung adenocarcinoma patients. MATERIAL AND METHODS: We included two datasets: 98 patients for discovery and 84 for validation. The phenotype of the primary tumor was quantified on pre-treatment CT-scans using 635 radiomic features. Univariate and multivariate analysis was performed to evaluate radiomics performance using the concordance index (CI). RESULTS: Thirty-five radiomic features were found to be prognostic (CI>0.60, FDR<5%) for DM and twelve for survival. It is noteworthy that tumor volume was only moderately prognostic for DM (CI=0.55, p-value=2.77×10(-5)) in the discovery cohort. A radiomic-signature had strong power for predicting DM in the independent validation dataset (CI=0.61, p-value=1.79×10(-17)). Adding this radiomic-signature to a clinical model resulted in a significant improvement of predicting DM in the validation dataset (p-value=1.56×10(-11)). CONCLUSIONS: Although only basic metrics are routinely quantified, this study shows that radiomic features capturing detailed information of the tumor phenotype can be used as a prognostic biomarker for clinically-relevant factors such as DM. Moreover, the radiomic-signature provided additional information to clinical data.

Outcomes by tumor histology and KRAS mutation status after lung stereotactic body radiation therapy for early-stage non-small-cell lung cancer.

BACKGROUND: We analyzed outcomes after lung stereotactic body radiotherapy (SBRT) for early-stage non-small cell lung-carcinoma (NSCLC) by histology and KRAS genotype. PATIENTS AND METHODS: We included 75 patients with 79 peripheral tumors treated with SBRT (18 Gy × 3 or 10 to 12 Gy × 5) at our institution from 2009 to 2012. Genotyping for KRAS mutations was performed in 10 patients. Outcomes were analyzed by the Kaplan-Meier method/Cox regression, or cumulative incidence method/Fine-Gray analysis. RESULTS: The median patient age was 74 (range, 46 to 93) years, and Eastern Cooperative Oncology Group performance status was 0 to 1 in 63%. Tumor histology included adenocarcinoma (44%), squamous cell carcinoma (25%), and NSCLC (18%). Most tumors were T1a (54%). Seven patients had KRAS-mutant tumors (9%). With a median follow-up of 18.8 months among survivors, the 1-year estimate of overall survival was 88%, cancer-specific survival (CSS) 92%, primary tumor control 94%, and freedom from recurrence (FFR) 67%. In patients with KRAS-mutant tumors, there was a significantly lower tumor control (67% vs. 96%; P = .04), FFR (48% vs. 69%; P = .03), and CSS (75% vs. 93%; P = .05). On multivariable analysis, histology was not associated with outcomes, but KRAS mutation (hazard ratio, 10.3; 95% confidence interval, 2.3-45.6; P = .0022) was associated with decreased CSS after adjusting for age. CONCLUSION: In this SBRT series, histology was not associated with outcomes, but KRAS mutation was associated with lower FFR on univariable analysis and decreased CSS on multivariable analysis. Because of the small sample size, these hypothesis-generating results need to be studied in larger data sets.

Low incidence of chest wall pain with a risk-adapted lung stereotactic body radiation therapy approach using three or five fractions based on chest wall dosimetry.

PURPOSE: To examine the frequency and potential of dose-volume predictors for chest wall (CW) toxicity (pain and/or rib fracture) for patients receiving lung stereotactic body radiotherapy (SBRT) using treatment planning methods to minimize CW dose and a risk-adapted fractionation scheme. METHODS: We reviewed data from 72 treatment plans, from 69 lung SBRT patients with at least one year of follow-up or CW toxicity, who were treated at our center between 2010 and 2013. Treatment plans were optimized to reduce CW dose and patients received a risk-adapted fractionation of 18 Gy×3 fractions (54 Gy total) if the CW V30 was less than 30 mL or 10-12 Gy×5 fractions (50-60 Gy total) otherwise. The association between CW toxicity and patient characteristics, treatment parameters and dose metrics, including biologically equivalent dose, were analyzed using logistic regression. RESULTS: With a median follow-up of 20 months, 6 (8.3%) patients developed CW pain including three (4.2%) grade 1, two (2.8%) grade 2 and one (1.4%) grade 3. Five (6.9%) patients developed rib fractures, one of which was symptomatic. No significant associations between CW toxicity and patient and dosimetric variables were identified on univariate nor multivariate analysis. CONCLUSIONS: Optimization of treatment plans to reduce CW dose and a risk-adapted fractionation strategy of three or five fractions based on the CW V30 resulted in a low incidence of CW toxicity. Under these conditions, none of the patient characteristics or dose metrics we examined appeared to be predictive of CW pain.