Proton beam therapy utilization in adults with primary brain tumors in the United States.

The utilization of proton beam therapy (PBT) as the primary treatment of adults with primary brain tumors (APBT) was evaluated through query of the National Cancer Database (NCDB) between the years 2004 and 2015. International Classification of Diseases for Oncology code for each patient was stratified into six histology categories; high-grade gliomas, medulloblastomas, ependymomas, other gliomas, other malignant tumors, or other benign intracranial tumors. Demographics of the treatment population were also analyzed. A total of 1,296 patients received PBT during the 11-year interval for treatment of their primary brain tumor. High-grade glioma, medulloblastoma, ependymoma, other glioma, other malignant, and other benign intracranial histologies made up 39%, 20%, 13%, 12%, 13%, and 2% of the cohort, respectively. The number of patients treated per year increased from 34 to 300 in years 2004 to 2015. Histologies treated with PBT varied over the 11-year interval with high-grade gliomas comprising 75% and 45% at years 2004 and 2015, respectively. The majority of the patient population was 18-29 years of age (59%), Caucasian race (73%), had median reported income of over $63,000 (46%), were privately insured (68%), and were treated at an academic institution (70%). This study characterizes trends of malignant and benign APBT histologies treated with PBT. Our data from 2004 through 2015 illustrates a marked increase in the utilization of PBT in the treatment of APBT and shows variability in the tumor histology treated over this time.

Quantitative assessment and clinical relevance of VEGFRs-positive tumor cells in refractory brain tumors.

Vascular endothelial growth factor (VEGF)/VEGF receptor (VEGFR)1 and 2 signaling is a potent activator of tumor angiogenesis. Although the expressions of VEGFR1 and VEGFR2 were initially thought to be limited to the endothelial cells, it is now known that both the receptors are expressed in tumor cells. This is the first study wherein VEGFRs-positive tumor cells are quantitatively evaluated for brain tumors with upregulated VEGF/VEGFR signaling. The percentage of VEGFRs-positive tumor cells was quantitatively evaluated in various brain tumors (10 glioblastomas, 22 neurofibromatosis type 2 [NF2]-related schwannomas, 21 sporadic schwannomas, 27 chordomas, 36 meningiomas, 29 hemangioblastomas, 11 hemangiopericytoma, and 13 ependymomas) using immunohistochemistry. VEGF-A expression was also analyzed using quantitative real-time polymerase chain reaction. Double immunofluorescence staining using anti-PDGFR-ß and anti-CD34 antibody, microvessel density, and vessel diameter were analyzed to evaluate the vascular characteristics. Chordomas demonstrated an extremely higher percentage of VEGFR1 and VEGFR2-positive tumor cells than other tumors. In contrast, meningiomas and hemangiopericytomas showed few VEGFRs-positive tumor cells. The percentage of positive tumor cells in chordomas, hemangioblastomas, and NF2 schwannomas was associated with clinical courses, such as shorter progression free survival, and growth speed. Glioblastomas and NF2 schwannomas showed larger tumor vessels without pericyte coverage. The present study is the first to quantitatively analyze VEGFR1- and VEGFR2- positive tumor cells in various types of refractory brain tumors. This novel parameter significantly correlated with the progressive clinical courses.

Molecular Subgroups of Glioblastoma- an Assessment by Immunohistochemical Markers.

Comprehensive molecular characterization of and novel therapeutic approaches to glioblastoma have been explored as a result of advancements in biotechnologies. In this study, we aimed to bring basic research discoveries closer to clinical practice and ultimately incorporate molecular classification into the routine histopathological evaluation of grade IV gliomas. Integrated results of genome-wide sequencing, transcriptomic and epigenomic analyses by The Cancer Genome Atlas Network defined the classic, proneural, neural and mesenchymal subtypes of this tumor. In a retrospective cohort, we analyzed selected subgroup-defining molecular markers in formalin-fixed paraffin-embedded surgical specimens by immunohistochemistry. Quantitative and qualitative scores of marker expression were tested in hierarchical cluster analyses to evaluate segregations of the molecular subgroups, which then were correlated with clinical parameters including patients' age, gender and overall survival. Our study has confirmed the separation of molecular glioblastoma subgroups with clear trends regarding clinical correlations. Future analyses in a larger, prospective cohort using similar methods are expected to facilitate the development of a molecular diagnostic panel that may complement routine histological work up and support prognostication as well as treatment decisions in glioblastoma.

Combining multimodal imaging and treatment features improves machine learning-based prognostic assessment in patients with glioblastoma multiforme.

BACKGROUND: For Glioblastoma (GBM), various prognostic nomograms have been proposed. This study aims to evaluate machine learning models to predict patients' overall survival (OS) and progression-free survival (PFS) on the basis of clinical, pathological, semantic MRI-based, and FET-PET/CT-derived information. Finally, the value of adding treatment features was evaluated. METHODS: One hundred and eighty-nine patients were retrospectively analyzed. We assessed clinical, pathological, and treatment information. The VASARI set of semantic imaging features was determined on MRIs. Metabolic information was retained from preoperative FET-PET/CT images. We generated multiple random survival forest prediction models on a patient training set and performed internal validation. Single feature class models were created including "clinical," "pathological," "MRI-based," and "FET-PET/CT-based" models, as well as combinations. Treatment features were combined with all other features. RESULTS: Of all single feature class models, the MRI-based model had the highest prediction performance on the validation set for OS (C-index: 0.61 [95% confidence interval: 0.51-0.72]) and PFS (C-index: 0.61 [0.50-0.72]). The combination of all features did increase performance above all single feature class models up to C-indices of 0.70 (0.59-0.84) and 0.68 (0.57-0.78) for OS and PFS, respectively. Adding treatment information further increased prognostic performance up to C-indices of 0.73 (0.62-0.84) and 0.71 (0.60-0.81) on the validation set for OS and PFS, respectively, allowing significant stratification of patient groups for OS. CONCLUSIONS: MRI-based features were the most relevant feature class for prognostic assessment. Combining clinical, pathological, and imaging information increased predictive power for OS and PFS. A further increase was achieved by adding treatment features.

Determination of optical properties of human brain tumor tissues from 350 to 1000 nm to investigate the cause of false negatives in fluorescence-guided resection with 5-aminolevulinic acid.

The optical properties of human brain tumor tissues, including glioblastoma, meningioma, oligodendroglioma, and metastasis, that were classified into "strong," "vague," and "unobservable" fluorescence by a neurosurgeon were measured and compared. The optical properties of the tissues were measured with a double integrating sphere and the inverse Monte Carlo technique from 350 to 1000 nm. Using reasons of ex-vivo measurement, the optical properties at around 420 nm were potentially affected by the hemoglobin content in tissues. Significant differences were not observed between the optical properties of the glioblastoma regions with "strong" and "unobservable" fluorescence. Sections of human brain tumor tissue with "strong" and "unobservable" fluorescence were stained with hematoxylin and eosin. The cell densities [mean ± standard deviation (S.D.)] in regions with "strong" and "unobservable" fluorescence were 31 ± 9 × 102 per mm2 and 12 ± 4 × 102 per mm2, respectively, which is a statistically significant difference. The higher fluorescence intensity is associated with higher cell density. The difference in cell density modified the scattering coefficient yet it does not lead to significant differences in the reduced scattering coefficient and thus does not affect the propagation of the diffuse fluorescent light. Hence, the false negatives, which mean a brain tumor only shows "unobservable" fluorescence and is hence classified incorrectly as nontumor, in using 5-ALA for detection of human glioblastoma do not result from the differences in optical properties of human brain glioblastoma tissues. Our results suggest that the primary cause of false negatives may be a lack of PpIX or a low accumulation of PpIX.

A simple algorithmic approach using histology and immunohistochemistry for the current classification of adult diffuse glioma in a resource-limited set-up.

AIMS: The WHO 2016 classification of diffuse gliomas combines histological and molecular parameters for diagnosis. However, in view of cost constraints for molecular testing, an economical working formula is essential to reach a meaningful diagnosis in a resource-limited setting. The aim of this study was to establish a practical algorithmic approach using histology and immunohistochemistry (IHC) in the classification of diffuse gliomas in such a set-up. METHODS: Diffuse gliomas of WHO grade II and III diagnosed in our institute in the year 2016 were analysed for histological and IHC features, using the markers isocitrate dehydrogenase 1 (IDH1R132H) and α thalassemia/mental retardation syndrome X-linked gene (ATRX). Fluorescence in situ hybridisation (FISH) for 1p/19q co-deletion was performed when requested. RESULTS: 449 diffuse gliomas (grades II/III) were included in the study. Integrating histology and IHC features, as per the WHO 2016 guidelines, we derived the following groups: Astrocytoma, IDH-mutant (A,IDH-mt, 37.2%); astrocytoma, not otherwise specified (A,NOS, 12.7%); oligoastrocytoma, NOS (OA,NOS, 4.5%); and oligodendroglioma, NOS (ODG,NOS, 45.6%). FISH was performed in a subset of ODG,NOS, OA,NOS and A,NOS gliomas. This revealed 1p/19q co-deletion in all cases of ODG,NOS, 15.8% of OA,NOS and 37.5% of A,NOS. Sequencing for rare IDH 1/2 mutations was not carried out in this study. CONCLUSION: In a resource-limited set-up, histology with IHC (IDH1(R132H) and ATRX) form the baseline to reasonably derive four histomolecular subgroups of diffuse glioma. Of these, we recommend, OA,NOS and IDH1(R132H)-non-mt ODG,NOS to be our priority for performing 1p/19q co-deletion studies in comparison to IDH-mt ODG,NOS, and it would not be mandatory for astrocytoma. Sequencing for rare IDH mutations is advised for A,NOS and OA,NOS groups, but not for the IDH1(R132H)-non-mutant diffuse gliomas with 1p/19q co-deletion.

Effect of Perfusion on Diffusion Kurtosis Imaging Estimates for In Vivo Assessment of Integrated 2016 WHO Glioma Grades : A Cross-Sectional Observational Study.

PURPOSE: To assess the role of perfusion-related signal decay on diffusion kurtosis imaging (DKI) estimates for in vivo stratification of glioma according to the integrated approach of the 2016 World Health Organization classification of tumors of the central nervous system (2016 CNS WHO). METHODS: In this study 77 patients with histopathologically confirmed glioma were retrospectively assessed between January 2013 and February 2017 in a prospective trial. Mean kurtosis (MK) and mean diffusivity (MD) metrics from DKI were assessed by two physicians blinded to the study from a volume of interest around the entire solid tumor. Wilcoxon's signed-rank test compared perfusion-biased and perfusion-corrected MK (MKpb and MKpc) and MD (MDpb, MDpc) values. One-way ANOVA was used to compare MKpb&pc and MDpb&pc values between 2016 WHO glioma grades. Spearman's correlation coefficient was used to correlate them with 2016 WHO glioma grades. Receiver operating characteristic (ROC) analysis was performed on MKpb&pc and MDpb&pc for the significant results. RESULTS: The MKpc values were significantly higher than MKpb values (p < 0.001), whereas MDpc values were significantly lower than MDpb values (p < 0.001). For stratifying gliomas, MKpb values (ROC AUC range, 0.818-0.979) showed a higher diagnostic performance than MKpc values (ROC AUC range, 0.773-0.975), whereas MDpb values (ROC AUC range, 0.744-0.928) showed less diagnostic performance than MDpc values (ROC AUC range, 0.753-0.934). The diagnostic accuracy of MKpb was 80.0%. CONCLUSION: The MK and MD estimates of DKI are influenced by microcapillary blood perfusion; however, taking the effect of perfusion on DKI metrics into account does not substantially impact their overall diagnostic performance in classifying glioma according to the 2016 CNS WHO.

Risk assessment in paediatric glioma-Time to move on from the binary classification.

BACKGROUND: Paediatric glioma encompasses a wide range of entities with highly variable prognoses. Gliomas are grouped by histopathological features into high- and low-grade glioma but this classification until recently has not taken into account many emerging risk factors in this disease. A comprehensive risk classification has not been published for paediatric glioma despite many risk factors being established in this disease. METHODS: A comprehensive literature review was carried out identifying risk factors for paediatric low-grade and high-grade glioma. RESULTS: The most consistently described risk factors in high-grade glioma included midline location and extent of surgical resection. For patients with progressive unresectable low-grade glioma, age under 1, neurofibromatosis type I status and location were the most consistently prognostic. Molecular classification shows promise in accurately reassigning diagnosis for some gliomas. CONCLUSION: Risk profiling in paediatric glioma will require a focused multinational effort but will result in a more accurate and nuanced assessment of prognosis.

Surgical assessment of the insula. Part 2: validation of the Berger-Sanai zone classification system for predicting extent of glioma resection.

OBJECTIVE: Though challenging, maximal safe resection of insular gliomas enhances overall and progression-free survival and deters malignant transformation. Previously published reports have shown that surgery can be performed with low morbidity. The authors previously described a Berger-Sanai zone classification system for insular gliomas. Using a subsequent dataset, they undertook this study to validate this zone classification system for predictability of extent of resection (EOR) in patients with insular gliomas. METHODS: The study population included adults who had undergone resection of WHO Grade II, III, or IV insular gliomas. In accordance with our prior published report, tumor location was classified according to the Berger-Sanai quadrant-style classification system into Zones I through IV. Interobserver variability was analyzed using a cohort of newly diagnosed insular gliomas and independent classification scores given by 3 neurosurgeons at various career stages. Glioma volumes were analyzed using FLAIR and T1-weighted contrast-enhanced MR images. RESULTS: One hundred twenty-nine procedures involving 114 consecutive patients were identified. The study population from the authors' previously published experience included 115 procedures involving 104 patients. Thus, the total experience included 244 procedures involving 218 patients with insular gliomas treated at the authors' institution. The most common presenting symptoms were seizure (68.2%) and asymptomatic recurrence (17.8%). WHO Grade II glioma histology was the most common (54.3%), followed by Grades III (34.1%) and IV (11.6%). The median tumor volume was 48.5 cm(3). The majority of insular gliomas were located in the anterior portion of the insula with 31.0% in Zone I, 10.9% in Zone IV, and 16.3% in Zones I+IV. The Berger-Sanai zone classification system was highly reliable, with a kappa coefficient of 0.857. The median EOR for all zones was 85%. Comparison of EOR between the current and prior series showed no change and Zone I gliomas continue to have the highest median EOR. Short- and long-term neurological complications remain low, and zone classification correlated with short-term complications, which were highest in Zone I and in Giant insular gliomas. CONCLUSIONS: The previously proposed Berger-Sanai classification system is highly reliable and predictive of insular glioma EOR and morbidity.

Glioma grade assessment by using histogram analysis of diffusion tensor imaging-derived maps.

INTRODUCTION: Current endeavors in neuro-oncology include morphological validation of imaging methods by histology, including molecular and immunohistochemical techniques. Diffusion tensor imaging (DTI) is an up-to-date methodology of intracranial diagnostics that has gained importance in studies of neoplasia. Our aim was to assess the feasibility of discriminant analysis applied to histograms of preoperative diffusion tensor imaging-derived images for the prediction of glioma grade validated by histomorphology. METHODS: Tumors of 40 consecutive patients included 13 grade II astrocytomas, seven oligoastrocytomas, six grade II oligodendrogliomas, three grade III oligoastrocytomas, and 11 glioblastoma multiformes. Preoperative DTI data comprised: unweighted (B (0)) images, fractional anisotropy, longitudinal and radial diffusivity maps, directionally averaged diffusion-weighted imaging, and trace images. Sampling consisted of generating histograms for gross tumor volumes; 25 histogram bins per scalar map were calculated. The histogram bins that allowed the most precise determination of low-grade (LG) or high-grade (HG) classification were selected by multivariate discriminant analysis. Accuracy of the model was defined by the success rate of the leave-one-out cross-validation. RESULTS: Statistical descriptors of voxel value distribution did not differ between LG and HG tumors and did not allow classification. The histogram model had 88.5% specificity and 85.7% sensitivity in the separation of LG and HG gliomas; specificity was improved when cases with oligodendroglial components were omitted. CONCLUSION: Constructing histograms of preoperative radiological images over the tumor volume allows representation of the grade and enables discrimination of LG and HG gliomas which has been confirmed by histopathology.

Computer-based assessment of cognitive functions in brain tumor patients.

NeuroCogFX is a short yet comprehensive computer-based neuropsychological battery of tests developed to investigate neurological patients for cognitive dysfunction after potentially neurotoxic therapy. NeuroCogFX had been standardized in a group of 242 healthy controls (Fliessbach et al., Fortschr Neurol Psychiatr 74:643-650, 2006). The present study was conducted to assess the practicability, reliability, and validity of NeuroCogFX in brain tumor patients without active disease after tumor-directed therapy. To evaluate its validity, neuropsychological testing with NeuroCogFX was completed parallel to a battery of established neuropsychological tests in 54 patients with different types of brain tumors and without active disease for at least 6 months. Retest reliability was assessed in a different sample of 49 patients with gliomas. Results showed good practicability with a median test duration of 28 min (range 16-51 min). Most subtests showed medium-sized retest reliability in healthy controls and tumor patients, with the exception of the 2-back test and reaction time measures in tumor patients. Convergent validity was confirmed for the domains psychomotor speed, verbal memory, and verbal short-term memory. NeuroCogFX enables serial scientific neuropsychological assessment of brain tumor patients. It can be carried out within a short period of time by non-academic personnel and is therefore applicable to large cohorts, e.g., within clinical trials.

Assessment of 31P-NMR analysis of phospholipid profiles for potential differential diagnosis of human cerebral tumors.

We describe a novel protocol for the non-histological diagnosis of human brain tumors in vitro combining high-resolution (31)P magnetic resonance spectroscopy ((31)P-MRS) of their phospholipid profile and statistical multivariate analysis. Chloroform/methanol extracts from 40 biopsies of human intracranial tumors obtained during neurosurgical procedures were prepared and analyzed by high-resolution (31)P-MRS. The samples were grouped in the following seven major classes: normal brain (n = 3), low-grade astrocytomas (n = 4), high-grade astrocytomas (n = 7), meningiomas (n = 9), schwannomas (n = 3), pituitary adenomas (n = 4), and metastatic tumors (n = 4). The phospholipid profile of every biopsy was determined by (31)P-NMR analysis of its chloroform/methanol extract and characterized by 19 variables including 10 individual phospholipid contributions and 9 phospholipid ratios. Most tumors depicted a decrease in phosphatidylethanolamine (PtdEtn) and phosphatidylserine (PtdSer), the former mainly in neuroepithelial neoplasms and the latter in metastases. An increase in phosphatidylcholine (PtdCho) and phosphatidylinositol (PtdIns) appeared predominantly in primary non-neuroepithelial tumors. Linear discriminant analysis (LDA) revealed the optimal combination of variables that could classify each biopsy between every pair of classes. The resultant discriminant functions were used to calculate the probability of correct classifications for each individual biopsy within the seven classes considered. Multilateral analysis classified correctly 100% of the normal brain samples, 89% of the meningiomas, 75% of the metastases, and 57% of the high-grade astrocytomas. The use of phospholipid profiles may complement appropriately previously proposed methods of intelligent diagnosis of human cerebral tumors.

Long-term prognostic assessment of 185 newly diagnosed gliomas: Grade III glioma showed prognosis comparable to that of Grade II glioma.

OBJECTIVE: We evaluated the prognoses of newly diagnosed gliomas through WHO Grades II, III and IV to assess the overall tendency of treatment results for glioma in our institute. Furthermore, statistical analysis was performed to determine factors influencing the prognosis. METHODS: A total of 185 newly diagnosed glioma patients were operated on from 2000 to 2006. The primary endpoint was the overall survival from the date of surgery. The factors assessed as to whether they influenced the prognosis were the WHO grades of sex, age, location of the lesion, pre-operative Karnofsky Performance Status (KPS), extent of resection and whether or not radiation therapy was performed. RESULTS: The WHO grades influenced the survival significantly (P < 0.0001). The Grades II and III showed no statistically significant difference in survival (P = 0.174), whereas Grades III and IV showed a significant difference (P < 0.0001). The factor influencing survival as well as the grades was the KPS (P < 0.0001). The comparison of survival over WHO grades in the same KPS group was performed for 2 KPS groups (KPS = 100, KPS 80-90), and these also showed significant differences (P = 0.0009 and 0.0143, respectively). CONCLUSIONS: Despite the different distributions of the KPS, the Grade III glioma patients showed survival comparable to that of the Grade II. On the other hand, the Grade IV glioma patients showed significantly poorer survival compared with Grade II or III.

Prospective assessment of activities of daily living using modified Barthel's Index in children and young adults with low-grade gliomas treated with stereotactic conformal radiotherapy.

PURPOSE: To report prospective evaluations of activities of daily living (ADL) in young patients with low-grade gliomas treated with stereotactic conformal radiotherapy (SCRT). MATERIALS AND METHODS: Between April 2001 and February 2008, 38 children and young adults (age 5-25 years, median 12.5 years) with low-grade gliomas with residual/progressive disease and treated with SCRT were accrued in a prospective protocol. Patients underwent baseline and follow-up ADL assessments by the modified Barthel's battery, which comprises domains of personal hygiene, bathing self, feeding, toilet, stair climbing, dressing, bowel control, bladder control, ambulation, and chair-bed transfer. RESULT: The patient population consisted of 38 patients (male 29, female 9) with a diagnosis of residual or progressive low-grade glioma (pilocytic astrocytoma in 27, fibrillary astrocytoma in 5, ependymoma in 4, and oligodendroglioma and pleomorphic xanthoastrocytoma in 1 each). Three patients were visually handicapped. Mean of total modified Barthel's ADL score (Barthel' Index, BI) at baseline before staring SCRT was 94.5 (standard deviation 14.8, range 45-100). At 2-year and 3-year follow-up, mean BI was 97.1 and 99, respectively. At baseline pre-radiotherapy assessment, patients with impaired visual function and with low performance status (Karnofsky performance score, KPS < 70) had significantly lower BI than those with normal vision (P

Effect of age and tumor grade on BOLD functional MR imaging in preoperative assessment of patients with glioma.

PURPOSE: To retrospectively determine if there was a combined effect of advanced age and increased tumor grade on blood oxygen level-dependent (BOLD) functional magnetic resonance (MR) imaging signal intensity (SI) in patients with gliomas. MATERIALS AND METHODS: The institutional review board approved this HIPAA-compliant study, and the informed consent requirement was waived. Data from 39 patients (27 men, 12 women; age range, 19-78 years) who had histopathologically confirmed gliomas and who underwent surgery after preoperative functional MR imaging were analyzed. Fourteen patients had grade II or grade III gliomas, and 25 patients had grade IV gliomas. A change in BOLD SI was measured in motor cortices of tumor-containing and non-tumor-containing hemispheres. The effect of age and tumor grade, both individually and together, on BOLD functional MR SI was assessed with t tests and regression analysis. RESULTS: In patients with grade IV gliomas, SI change was lower in the tumor-containing hemisphere than in the non-tumor-containing hemisphere (P = .012). SI change decreased with increased age in the tumor-containing hemisphere in patients with grade II or III gliomas (P = .032) and in the non-tumor-containing hemisphere in patients with grade IV gliomas (P = .026). While advanced age and increased glioma grade reduced SI change, the combined effect of these factors was not additive. In patients with grade IV gliomas, tumor presence reduced SI change, but the level of reduction was uniform across all ages and did not correlate with age (P = .541). CONCLUSION: In older patients with grade IV gliomas, BOLD SI is equivalent to that measured in younger patients with grade IV gliomas. Advanced age and tumor grade do not have a combined effect for reduction of BOLD SI. Rather, in patients with grade IV gliomas, tumor grade played a dominant role in reduction of SI change, whereas in patients with grade II and III gliomas, reduction of SI change correlated with only advanced age.

MRS quality assessment in a multicentre study on MRS-based classification of brain tumours.

This paper reports on quality assessment of MRS in the European Union-funded multicentre project INTERPRET (International Network for Pattern Recognition of Tumours Using Magnetic Resonance; http://azizu.uab.es/INTERPRET), which has developed brain tumour classification software using in vivo proton MR spectra. The quality assessment consisted of both MR system quality assurance (SQA) and quality control (QC) of spectral data acquired from patients and healthy volunteers. The system performance of the MR spectrometers at all participating centres was checked bimonthly by a short measurement protocol using a specially designed INTERPRET phantom. In addition, a more extended SQA protocol was performed yearly and after each hardware or software upgrade. To compare the system performance for in vivo measurements, each centre acquired MR spectra from the brain of five healthy volunteers. All MR systems fulfilled generally accepted minimal system performance for brain MRS during the entire data acquisition period. The QC procedure of the MR spectra in the database comprised automatic determination of the signal-to-noise ratio (SNR) in a water-suppressed spectrum and of the line width of the water resonance (water band width, WBW) in the corresponding non-suppressed spectrum. Values of SNR > 10 and WBW < 8 Hz at 1.5 T were determined empirically as conservative threshold levels required for spectra to be of acceptable quality. These thresholds only hold for SNR and WBW values using the definitions and data processing described in this article. A final QC check consisted of visual inspection of each clinically validated water-suppressed metabolite spectrum by two, or, in the case of disagreement, three, experienced MR spectroscopists, to detect artefacts such as large baseline distortions, exceptionally broadened metabolite peaks, insufficient removal of the water line, large phase errors, and signals originating from outside the voxel. In the end, 10% of 889 spectra with completed spectroscopic judgement were discarded.

Causal probabilistic modeling for malignancy grading in pathology with explanations of dependency to the related histological features.

This work demonstrates that histological grading of brain tumors and astrocytomas can be accurately predicted and causally explained with the help of causal probabilistic models, also known as Bayesian networks (BN). Although created statistically, this allows individual identification of the grade of malignancy as an internal cause that has enabled the development of the histological features to their observed state. The BN models are built from data representing 794 cases of astrocytomas with their malignant grading and corresponding histological features. The computerized learning process is improved when pre-specified knowledge (from the pathologist) about simple dependency relations to the histological features is taken into account. We use the BN models for both grading and causal analysis. In addition, the BN models provide a causal explanation of dependency between the histological features and the grading. This can offer the biggest potential for choice of an efficient treatment, since it concentrates on the malignancy grade as the cause of pathological observations. The causal analysis shows that all ten histological features are important for the grading. The histological features are causally ordered, implying that features of first order are of higher priority, e.g. for the choice of treatment in order not to allow the malignancy to progress to a higher degree. Due to the explanations of feature relations, the causal analysis can be considered as a powerful complement to any malignancy classification tool and allows reproducible comparison of malignancy grading.