Search

Background: Combining MRI techniques with machine learning methodology is rapidly gaining attention as a promising method for staging of brain gliomas. This study assesses the diagnostic value of such a framework applied to dynamic susceptibility contrast (DSC)-MRI in classifying treatment-naïve gliomas from a multi-center patients into WHO grades II-IV and across their isocitrate dehydrogenase (IDH) mutation status. Methods: Three hundred thirty-three patients from 6 tertiary centres, diagnosed histologically and molecularly with primary gliomas (IDH-mutant = 151 or IDH-wildtype = 182) were retrospectively identified. Raw DSC-MRI data was post-processed for normalised leakage-corrected relative cerebral blood volume (rCBV) maps. Shape, intensity distribution (histogram) and rotational invariant Haralick texture features over the tumour mask were extracted. Differences in extracted features across glioma grades and mutation status were tested using the Wilcoxon two-sample test. A random-forest algorithm was employed (2-fold cross-validation, 250 repeats) to predict grades or mutation status using the extracted features. Results: Shape, distribution and texture features showed significant differences across mutation status. WHO grade II-III differentiation was mostly driven by shape features while texture and intensity feature were more relevant for the III-IV separation. Increased number of features became significant when differentiating grades further apart from one another. Gliomas were correctly stratified by mutation status in 71% and by grade in 53% of the cases (87% of the gliomas grades predicted with distance less than 1). Conclusions: Despite large heterogeneity in the multi-center dataset, machine learning assisted DSC-MRI radiomics hold potential to address the inherent variability and presents a promising approach for non-invasive glioma molecular subtyping and grading. © 2020 The Author(s).

To investigate, along the pituitary- testis- prostate axis, the potential of preoperative serum TT in contributing to defining separate prostatectomy Gleason score (pGS) groups of the prostate cancer (PC) population.The data of 126 patients operated on for PC were retrospectively reviewed. No patient had previously received 5α-reductase inhibitor, luteinizing hormone (LH)-releasing hormone analogs or testosterone replacement treatment. The patient population was grouped according to the prostatectomy Gleason score (pGS) as 6=3+3, 7=3+4, 7=4+3 and 8-10. Twelve variables were simultaneously investigated in each group: age, prolactin (PRL), follicle stimulating hormone (FSH), LH, total testosterone (TT), free testosterone (FT), estradiol (Er), prostate specific antigen (PSA), percentage of prostate biopsy positive cores (P+), biopsy Gleason score (bGS), overall cancer volume estimated as percentage of prostate volume (V+) and prostate weight (Wi). Univariate analysis of variance (ANOVA), multivariate analysis of variance (MANOVA) and multivariate discriminant analysis were the statistical methods used for evaluating the data.There were 38 patients in pGS 6=3+3, 57 in pGS 7=3+4, 15 in pGS 7=4+3 and 16 in pGS 8-10. ANOVA showed that bGS (p0.0001), P+ (p0.0001), V+ (p0.0001), PSA (p=0.02), Wi (p=0.001) and TT (p=0.04) were significantly different in the four pGS groups. MANOVA tests showed that only bGS (p0.0001) and TT (p=0.005) were the significant variables that individually and independently contributed a significant amount to separation of the four pGS groups of the PC population. Multivariate discriminant analysis confirmed that TT (p=0.005) and bGS (p0.0001) were the only variables that independently and significantly contributed to separating the pGS groups.along the pituitary- testis- prostate axis, serum TT is a significant preoperative variable that independently contributes to separating the prostate cancer population into pGS score groups. Pretreatment baseline serum TT levels should be measured for their inclusion in nomograms and future neural networks to be considered in the patient population diagnosed with PC.

To evaluate Luteinizing hormone (LH) physiopathology along the pituitary testicular prostate axis at the time of initial diagnosis of prostate cancer in relation to the available clinical variables and to the subsequent cluster selection of the patient population.Age, percentages of positive cores at Trans Rectal Ultrasound Scan Biopsy (TRUSB) (P+), biopsy Gleason score (bGS), LH, Total Testosterone (TT), Free Testosterone (FT) and Prostate Specific Antigen (PSA) were the continuous clinical variables. All patients had histologically proven carcinoma of the prostate and had not previously received 5α-reductase inhibitors, LH-releasing hormone analogues or testosterone replacement treatment. Correlation analysis was performed for the patient population. Correlation analysis, linear regression and analysis of variance was computed in groups and subgroups of the prostate cancer population.Correlation analysis of the patient population showed that LH was significantly correlated to age (p=0.02) and FT (p=0.01). The population was clustered in LH I (LH≤7.5 IU/l) and LH II (LH7.5 IU/l). Correlation analysis showed significant LH correlations for TT (p0.0001) and FT (p=0.0004) for LH I; significant LH correlation to FT (p=0.0001) for LH II. Simple linear regression showed that LH was significantly predicted by both TT (p-Value0.0001) and FT (p-Value=0.0004) in LH I; but only FT (p-Value0.0001) in LH II. Multiple linear regression showed that LH was significantly predicted by both TT (p-Value=0.0004) and PSA (p-Value=0.03) in LH I; but only by FT (p-Value=0.003) in LH II. Analysis of variance showed that: a) LH and age were significantly lower in LH I than II; b) LH I expressed higher mean FT levels (p=0.08) and lower mean P+ (p=0.07) than LH II. The LH versus PSA plot was computed for LH group I and 3 sub clusters were created: LH I group A (LH/PSA≤0.25), B (0.25LH/PSA ≤0.75), and C (LH/PSA0.75). Correlation analysis showed that LH was significantly correlated to age (p=0.01), TT (p=0.03) and PSA (p=0.0004) in LH IA; LH was significantly correlated to PSA (p0.0001) in LH IB; and LH significantly correlated to TT (p=0.005), FT (p=0.01), and PSA (p=0.008) in LH 1C. Multiple linear regression showed that LH was significantly correlated to age (p=0.02) and PSA (p=0.01) in LH IA, to TT (p=0.01) and PSA (p0.0001) in LH IB, and to PSA (p=0.003) and weakly to TT (p=0.09) in LH IC. The groups differed significantly for mean levels of LH (p=0.0004), TT (p=0.005), FT (p=0.01), PSA (p0.0001), bGS (p=0.003). Analysis of variance between the subgroups of the patient population (LH IA, LH IB, LH IC, LH II) showed significant differences in mean levels for LH (p0.0001), age (p=0.004), TT (p=0.009), FT (p=0.02), PSA (p0.0001), PSA/FT (p0.0001), bGS (p=0.01), but not for P+ (p=0.10).According to LH physiopathology, the prostate cancer population could be clustered into hypo-gonadic and non-hypo-gonadic group at diagnosis. The hypo-gonadic group expresses an aggressive tumor phenotype and might be divided into two more different significant subsets including primary and secondary hypo-gonadic patients: the former (LH II) including older patients with high LH levels, the latter (LH IA) including younger patients with low LH and LH/PSA levels (subgroup LH IA). The non-hypo-gonadic group showed a less aggressive tumor phenotype and according to the LH/PSA ratio might beclustered into LH IB (0.25LH/PSA≤0.75) and LH IC (LH/PSA0.75), the former showing a more aggressive tumor phenotype than the latter. Confirmatory studies are necessary.