Machine Learning in Differentiating Gliomas from Primary CNS Lymphomas: A Systematic Review, Reporting Quality, and Risk of Bias Assessment.

BACKGROUND: Differentiating gliomas and primary CNS lymphoma represents a diagnostic challenge with important therapeutic ramifications. Biopsy is the preferred method of diagnosis, while MR imaging in conjunction with machine learning has shown promising results in differentiating these tumors. PURPOSE: Our aim was to evaluate the quality of reporting and risk of bias, assess data bases with which the machine learning classification algorithms were developed, the algorithms themselves, and their performance. DATA SOURCES: Ovid EMBASE, Ovid MEDLINE, Cochrane Central Register of Controlled Trials, and the Web of Science Core Collection were searched according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. STUDY SELECTION: From 11,727 studies, 23 peer-reviewed studies used machine learning to differentiate primary CNS lymphoma from gliomas in 2276 patients. DATA ANALYSIS: Characteristics of data sets and machine learning algorithms were extracted. A meta-analysis on a subset of studies was performed. Reporting quality and risk of bias were assessed using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) and Prediction Model Study Risk Of Bias Assessment Tool. DATA SYNTHESIS: The highest area under the receiver operating characteristic curve (0.961) and accuracy (91.2%) in external validation were achieved by logistic regression and support vector machines models using conventional radiomic features. Meta-analysis of machine learning classifiers using these features yielded a mean area under the receiver operating characteristic curve of 0.944 (95% CI, 0.898-0.99). The median TRIPOD score was 51.7%. The risk of bias was high for 16 studies. LIMITATIONS: Exclusion of abstracts decreased the sensitivity in evaluating all published studies. Meta-analysis had high heterogeneity. CONCLUSIONS: Machine learning-based methods of differentiating primary CNS lymphoma from gliomas have shown great potential, but most studies lack large, balanced data sets and external validation. Assessment of the studies identified multiple deficiencies in reporting quality and risk of bias. These factors reduce the generalizability and reproducibility of the findings.

Diagnostic Accuracy of T1-Weighted Dynamic Contrast-Enhanced-MRI and DWI-ADC for Differentiation of Glioblastoma and Primary CNS Lymphoma.

BACKGROUND AND PURPOSE: Glioblastoma and primary CNS lymphoma dictate different neurosurgical strategies; it is critical to distinguish them preoperatively. However, current imaging modalities do not effectively differentiate them. We aimed to examine the use of DWI and T1-weighted dynamic contrast-enhanced-MR imaging as potential discriminative tools. MATERIALS AND METHODS: We retrospectively reviewed 18 patients with primary CNS lymphoma and 36 matched patients with glioblastoma with pretreatment DWI and dynamic contrast-enhanced-MR imaging. VOIs were drawn around the tumor on contrast-enhanced T1WI and FLAIR images; these images were transferred onto coregistered ADC maps to obtain the ADC and onto dynamic contrast-enhanced perfusion maps to obtain the plasma volume and permeability transfer constant. Histogram analysis was performed to determine the mean and relative ADCmean and relative 90th percentile values for plasma volume and the permeability transfer constant. Nonparametric tests were used to assess differences, and receiver operating characteristic analysis was performed for optimal threshold calculations. RESULTS: The enhancing component of primary CNS lymphoma was found to have significantly lower ADCmean (1.1 × 10-3 versus 1.4 × 10-3; P < .001) and relative ADCmean (1.5 versus 1.9; P < .001) and relative 90th percentile values for plasma volume (3.7 versus 5.0; P < .05) than the enhancing component of glioblastoma, but not significantly different relative 90th percentile values for the permeability transfer constant (5.4 versus 4.4; P = .83). The nonenhancing portions of glioblastoma and primary CNS lymphoma did not differ in these parameters. On the basis of receiver operating characteristic analysis, mean ADC provided the best threshold (area under the curve = 0.83) to distinguish primary CNS lymphoma from glioblastoma, which was not improved with normalized ADC or the addition of perfusion parameters. CONCLUSIONS: ADC was superior to dynamic contrast-enhanced-MR imaging perfusion, alone or in combination, in differentiating primary CNS lymphoma from glioblastoma.

First-line treatment and outcome of elderly patients with primary central nervous system lymphoma (PCNSL)--a systematic review and individual patient data meta-analysis.

BACKGROUND: To investigate prognosis and effects of first-line therapy in elderly primary central nervous system lymphoma (PCNSL) patients. PATIENTS AND METHODS: A systematic review of studies about first-line therapy in immunocompetent patients ≥60 years with PCNSL until 2014 and a meta-analysis of individual patient data from eligible studies and international collaborators were carried out. RESULTS: We identified 20 eligible studies; from 13 studies, we obtained individual data of 405 patients, which were pooled with data of 378 additional patients (N = 783). Median age and Karnofsky Performance Score (KPS) was 68 years (range: 60-90 years) and 60% (range: 10%-100%), respectively. Treatments varied greatly, 573 (73%) patients received high-dose methotrexate (HD-MTX)-based therapy. A total of 276 patients received whole-brain radiotherapy (median 36 Gy, range 28.5-70 Gy). KPS ≥ 70% was the strongest prognostic factor for mortality [hazard ratio (HR) 0.50, 95% confidence interval (CI) 0.41-0.62]. After a median follow-up of 40 months, HD-MTX-based therapy was associated with improved survival (HR 0.70, 95% CI 0.53-0.93). There was no difference between HD-MTX plus oral chemotherapy and more aggressive HD-MTX-based therapies (HR 1.39, 95% CI 0.90-2.15). Radiotherapy was associated with an improved survival, but correlated with an increased risk for neurological side-effects (odds ratio 5.23, 95% CI 2.33-11.74). CONCLUSIONS: Elderly PCNSL patients benefit from HD-MTX-based therapy, especially if combined with oral alkylating agents. More aggressive HD-MTX protocols do not seem to improve outcome. WBRT may improve outcome, but is associated with increased risk for neurological side-effects. Prospective trials for elderly PCNSL patients are warranted.

Diffusion and perfusion MRI to differentiate treatment-related changes including pseudoprogression from recurrent tumors in high-grade gliomas with histopathologic evidence.

BACKGROUND AND PURPOSE: Treatment-related changes and recurrent tumors often have overlapping features on conventional MR imaging. The purpose of this study was to assess the utility of DWI and DSC perfusion imaging alone and in combination to differentiate treatment-related effects and recurrent high-grade gliomas. MATERIALS AND METHODS: We retrospectively identified 68 consecutive patients with high-grade gliomas treated by surgical resection followed by radiation therapy and temozolomide, who then developed increasing enhancing mass lesions indeterminate for treatment-related changes versus recurrent tumor. All lesions were diagnosed by histopathology at repeat surgical resection. ROI analysis was performed of the enhancing lesion on the ADC and DSC maps. Measurements made by a 2D ROI of the enhancing lesion on a single slice were recorded as ADCLesion and rCBVLesion, and measurements made by the most abnormal small fixed diameter ROI as ADCROI and rCBVROI. Statistical analysis was performed with Wilcoxon rank sum tests with P = .05. RESULTS: Ten of the 68 patients (14.7%) had treatment-related changes, while 58 patients (85.3%) had recurrent tumor only (n = 19) or recurrent tumor mixed with treatment effect (n = 39). DWI analysis showed higher ADCLesion in treatment-related changes than in recurrent tumor (P = .003). DSC analysis revealed lower relative cerebral blood volume (rCBV)Lesion and rCBVROI in treatment-related changes (P = .003 and P = .011, respectively). Subanalysis of patients with suspected pseudoprogression also revealed higher ADCLesion (P = .001) and lower rCBVLesion (P = .028) and rCBVROI (P = .032) in treatment-related changes. Applying a combined ADCLesion and rCBVLesion model did not outperform either the ADC or rCBV metric alone. CONCLUSIONS: Treatment-related changes showed higher diffusion and lower perfusion than recurrent tumor. Similar correlations were found for patients with suspected pseudoprogression.

Pretreatment Dynamic Susceptibility Contrast MRI Perfusion in Glioblastoma: Prediction of EGFR Gene Amplification.

BACKGROUND AND PURPOSE: Molecular and genetic testing is becoming increasingly relevant in GBM. We sought to determine whether dynamic susceptibility contrast (DSC) magnetic resonance imaging (MRI) perfusion imaging could predict EGFR-defined subtypes of GBM. MATERIALS AND METHODS: We retrospectively identified 106 consecutive glioblastoma (GBM) patients with known EGFR gene amplification, and a subset of 65 patients who also had known EGFRvIII gene mutation status. All patients underwent T2* DSC MRI perfusion. DSC perfusion maps and T2* signal intensity time curves were evaluated, and the following measures of tumor perfusion were recorded: (1) maximum relative cerebral blood volume (rCBV), (2) relative peak height (rPH), and (3) percent signal recovery (PSR). The imaging metrics were correlated to EGFR gene amplification and EGFRvIII mutation status using univariate analyses. RESULTS: EGFR amplification was present in 44 (41.5 %) subjects and absent in 62 (58.5 %). Among the 65 subjects who had undergone EGFRvIII mutation transcript analysis, 18 subjects (27.7 %) tested positive for the EGFRvIII mutation, whereas 47 (72.3 %) did not. Higher median rCBV (3.31 versus 2.62, p = 0.01) and lower PSR (0.70 versus 0.78, p = 0.03) were associated with high levels of EGFR amplification. Higher median rPH (3.68 versus 2.76, p = 0.03) was associated with EGFRvIII mutation. CONCLUSION: DSC MRI perfusion may have a role in identifying patients with EGFR gene amplification and EGFRvIII gene mutation status, potential targets for individualized treatment protocols. Our results raise the need for further investigation for imaging biomarkers of genetically unique GBM subtypes.

Potential role of preoperative conventional MRI including diffusion measurements in assessing epidermal growth factor receptor gene amplification status in patients with glioblastoma.

BACKGROUND AND PURPOSE: Epidermal growth factor receptor amplification is a common molecular event in glioblastomas. The purpose of this study was to examine the potential usefulness of morphologic and diffusion MR imaging signs in the prediction of epidermal growth factor receptor gene amplification status in patients with glioblastoma. MATERIALS AND METHODS: We analyzed pretreatment MR imaging scans from 147 consecutive patients with newly diagnosed glioblastoma and correlated MR imaging features with tumor epidermal growth factor receptor amplification status. The following morphologic tumor MR imaging features were qualitatively assessed: 1) border sharpness, 2) cystic/necrotic change, 3) hemorrhage, 4) T2-isointense signal, 5) restricted water diffusion, 6) nodular enhancement, 7) subependymal enhancement, and 8) multifocal discontinuous enhancement. A total of 142 patients had DWI available for quantitative analysis. ADC maps were calculated, and the ADCmean, ADCmin, ADCmax, ADCROI, and ADCratio were measured. RESULTS: Epidermal growth factor receptor amplification was present in 60 patients (40.8%) and absent in 87 patients (59.2%). Restricted water diffusion correlated with epidermal growth factor receptor amplification (P = .04), whereas the other 7 morphologic MR imaging signs did not (P > .12). Quantitative DWI analysis found that all ADC measurements correlated with epidermal growth factor receptor amplification, with the highest correlations found with ADCROI (P = .0003) and ADCmean (P = .0007). CONCLUSIONS: Our results suggest a role for diffusion MR imaging in the determination of epidermal growth factor receptor amplification status in glioblastoma. Additional work is necessary to confirm these results and isolate new imaging biomarkers capable of noninvasively characterizing the molecular status of these tumors.

Potential utility of conventional MRI signs in diagnosing pseudoprogression in glioblastoma.

OBJECTIVE: To examine the potential utility of conventional MRI signs in differentiating pseudoprogression (PsP) from early progression (EP). METHODS: This retrospective study reviewed initial postradiotherapy MRI scans of 321 patients with glioblastoma undergoing chemotherapy and radiotherapy. A total of 93 patients were found to have new or increased enhancing mass lesions, raising the possibility of PsP. Final diagnosis of PsP or EP was established upon review of surgical specimens from a second resection or by clinical and radiologic follow-up. A total of 11 MRI signs potentially helpful in the differentiation between PsP and EP were examined on the initial post-RT MRI and were correlated with the final diagnosis through χ(2) or Fisher exact test. RESULTS: Sixty-three (67.7%) of the 93 patients had EP, of which 22 (34.9%) were diagnosed by pathology. Thirty patients (32.3%) had PsP; 6 (16.7% of the 30) were diagnosed by pathology. Subependymal enhancement was predictive for EP (p = 0.001) with 38.1% sensitivity, 93.3% specificity, and 41.8% negative predictive value. The other 10 signs had no predictive value (p = 0.06-1.0). CONCLUSIONS: Conventional MRI signs have limited utility in diagnosing PsP in patients with recently treated glioblastomas and worsening enhancing lesions. We did not find a sign with a high negative predictive value for PsP that would have been the most useful for the clinical physician. When present, subependymal spread of the enhancing lesion is a useful MRI marker in identifying EP rather than PsP.

Primary CNS lymphoma with intraocular involvement: International PCNSL Collaborative Group Report.

OBJECTIVE: To describe the demographics, diagnostic details, therapeutic management, and outcome in patients with primary CNS lymphoma (PCNSL) with ocular involvement. METHODS: A retrospective study of 221 patients was assembled from 16 centers in seven countries. Only HIV-negative, immunocompetent patients with brain and ocular lymphoma were included; none had systemic lymphoma. RESULTS: Median age at diagnosis was 60. Fifty-seven percent were women. Median Eastern Cooperative Oncology Group performance status was 2. Ocular disturbance and behavioral/cognitive changes were the most common presenting symptoms. Diagnosis of lymphoma was made by brain biopsy (147), vitrectomy (65), or CSF cytology (11). Diagnosis of intraocular lymphoma was made by vitrectomy/choroidal/retinal biopsy (90) or clinical ophthalmic examination (141). CSF cytology was positive in 23%. Treatment information was available for 176 patients. A total of 102 received dedicated ocular therapy (ocular radiotherapy 79, intravitreal methotrexate 22, and both 1) in addition to treatment for their brain lymphoma. Sixty-nine percent progressed at a median of 13 months; sites of progression included brain 52%, eyes 19%, brain and eyes 12%, and systemic 2%. Patients treated with local ocular therapy did not have a statistically significant decreased risk of failing in the eyes (p = 0.7). Median progression free survival and overall survival for the entire cohort were 18 and 31 months. CONCLUSION: This is the largest reported series of primary CNS lymphoma (PCNSL) with intraocular involvement. Progression free and overall survival was similar to that reported with PCNSL. Dedicated ocular therapy improved disease control but did not affect overall survival.

Primary intraocular lymphoma: an International Primary Central Nervous System Lymphoma Collaborative Group Report.

BACKGROUND: Primary intraocular lymphoma (PIOL) is an uncommon subset of primary central nervous system lymphoma. Because it is rare and difficult to diagnose, the natural history and optimal management are unknown. PATIENTS AND METHODS: A retrospective study of 83 HIV negative, immunocompetent PIOL patients was assembled from 16 centers in seven countries. RESULTS: Median age at diagnosis was 65. Median ECOG performance status was 0. Presenting symptoms included blurred vision, decreased visual acuity, and floaters. Median time to diagnosis was 6 months. Diagnosis was made by vitrectomy (74), choroidal/retinal biopsy (6) and ophthalmic exam (3). Eleven percent had positive CSF cytology. Initial treatment was categorized as focal in 23 (intra-ocular methotrexate, ocular radiotherapy) or extensive in 53 (systemic chemotherapy, whole brain radiotherapy). Six received none; details are unknown in one. Forty-seven relapsed: brain 47%, eyes 30%, brain and eyes 15%, and systemic 8%. Median time to relapse was 19 months. Focal therapy alone did not increase risk of brain relapse. Median progression free (PFS) and overall survival (OS) were 29.6 and 58 months, respectively, and unaffected by treatment type. CONCLUSION: Treatment type did not affect relapse pattern, median PFS or OS. Focal therapy may minimize treatment toxicity without compromising disease control.

Temozolomide for low-grade gliomas: predictive impact of 1p/19q loss on response and outcome.

OBJECTIVE: To evaluate the predictive impact of chromosome 1p/19q deletions on the response and outcome of progressive low-grade gliomas (LGG) treated with up-front temozolomide (TMZ) chemotherapy. METHODS: Adult patients with measurable, progressive LGG (WHO grade II) treated with TMZ delivered at the conventional schedule (200 mg/m(2)/day for 5 consecutive days, repeated every 28 days) were retrospectively evaluated for response by central review of MRI-s. Chromosome 1p and 19q deletions were detected by the loss of the heterozygosity technique (LOH). RESULTS: A total of 149 consecutive patients were included in this retrospective, single center observational study. The median number of TMZ cycles delivered was 14 (range 2 to 30). Seventy-seven patients (53%) experienced an objective response (including 22 [15%] cases of partial response and 55 [38%] cases of minor response), 55 (37%) patients had stable disease, and 14 (10%) had a progressive disease. The median time to maximum tumor response was 12 months (range 3 to 30 months). The median progression-free survival (PFS) was 28 months (95% CI: 23.4 to 32.6). Material for genotyping was available for 86 patients. Combined 1p/19q LOH was present in 42% of the cases and was significantly associated with a higher rate (p = 0.02) and longer objective response to chemotherapy (p = 0.017), and both longer PFS (p = 4.10(-5)) and overall survival (p = 0.04). CONCLUSION: Low-grade gliomas respond to temozolomide and loss of chromosome 1p/19q predicts both a durable chemosensitivity and a favorable outcome.

Vinorelbine combined with a protracted course of temozolomide for recurrent brain metastases: a phase I trial.

Temozolomide (TMZ) has shown modest efficacy in the treatment of recurrent brain metastasis (BM). We designed a new regimen utilizing dose-intensified, protracted course of TMZ in combination with vinorelbine, a lipophilic large-spectrum agent, in an attempt to improve the efficacy of TMZ. This phase I study was conducted to establish the maximum tolerated dose (MTD) of vinorelbine for this combination. Patients with recurrent or progressive BM were eligible. Chemotherapy consisted of 28-day cycles with TMZ (150 mg/m2, days 1-7 and 15-21) and vinorelbine (days one and eight at escalating doses). The starting dose was 15 mg/m2, with increments of 5 mg/m2 for each cohort of 3-6 patients, until MTD was reached (30 mg/m2). A total of 21 patients were enrolled; the median age was 59 (41-77). The primary tumor was lung cancer in 13 patients (NSCLC in 10, SCLC in 3), breast in 6, renal in 1 and endometrial in 1. Vinorelbine dose was 15 mg/m2 in seven patients, 20 mg/m2 in five, 25 mg/m2 in four and 30 mg/m2 in six. Grades 3 and 4 neutropenia developed in six patients, lymphopenia in nine, and thrombocytopenia in six; other toxicities were rare. No dose-limiting toxicity was seen. Out of 18 evaluable patients 2 had a radiographic response (one partial and one minor). Disease was stable in 6 of 18 patients and the median survival was 27 weeks. This regimen was well tolerated and a phase II trial using a dose of 30 mg/m2 of vinorelbine is warranted.

EGFR tyrosine kinase domain mutations in human gliomas.

Gefitinib is an epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor effective in patients with lung cancer with mutations in exons 19 and 21 of the EGFR tyrosine kinase domain. In this study, the authors tested the presence of such mutations in 95 gliomas including glioblastomas, anaplastic oligodendrogliomas, and low-grade gliomas. No mutation was found, which suggests that the biology of EGFR in gliomas is different from lung cancer and that this may be a factor in the resistance of glioblastomas to gefitinib.