Search

Background and purpose Advanced imaging analysis for the prediction of tumor biology and modelling of clinically relevant parameters using computed imaging features is part of the emerging field of radiomics research. Here we test the hypothesis that a machine learning approach can distinguish grade 1 from higher gradings in meningioma patients using radiomics features derived from a heterogenous multicenter dataset of multi-paramedic MRI. Methods A total of 138 patients from 5 international centers that underwent MRI prior to surgical resection of intracranial meningiomas were included. Segmentation was performed manually on co-registered multi-parametric MR images using apparent diffusion coefficient (ADC) maps, T1-weighted (T1), post-contrast T1-weighted (T1c), subtraction maps (Sub, T1c – T1), T2-weighted fluid-attenuated inversion recovery (FLAIR) and T2-weighted (T2) images. Feature selection was performed and using cross-validation to separate training from testing data, four machine learning classifiers were scored on combinations of MRI modalities: random forest (RF), extreme gradient boosting (XGBoost), support vector machine (SVM) and multilayer perceptron (MLP). Results The best AUC of 0.97 (1.0 and 0.97 for sensitivity and specificity) was observed for the combination of ADC, ADC of the peritumoral edema, T1, T1c, Sub and FLAIR-derived features using only 16 of the 10,914 possible features and XGBoost. Conclusions Machine learning using radiomics features derived from multi-parametric MRI is capable of high AUC scores with high sensitivity and specificity in classifying meningiomas between low and higher gradings despite heterogeneous protocols across different centers. Feature selection can be performed effectively even when extracting a large amount of data for radiomics fingerprinting.

Background and objectives Texture analysis derived from morphological magnetic resonance (MR) images might be associated with histopathology in tumors. The present study sought to elucidate possible associations between texture features derived from T1-and T2-weighted images with proliferation index Ki67 in soft tissue sarcomas. Methods Overall, 29 patients (n = 13, 44.8% female) with a median age of 52 years were included into this retrospective study. Several soft tissue sarcomas were investigated. Texture analysis was performed on pre-contrast T1-weighted and T2-weighted images using the free available Mazda software. Results The best correlation coefficients with Ki67 index were identified for the following parameters: T1-weighted images “45dgr_RLNonUni (p = 0.50, P = 0.006), T2-weighted images “S (4,0)SumAverg” (p = −0.45, P = 0.02). A ROC analysis was performed for Ki67-index with a threshold of 10%. The highest area under the curve (AUC) was found for the parameter “T1_WavEnHL_s-7” with an AUC of 0.90. For the threshold of Ki67 = 20% the highest AUC was identified for the parameter „T2_S (1,1)Entropy” with an AUC of 0.77. Conclusion Several texture features derived from T1-and T2-weighted images correlated with proliferation index Ki67 and might be used as valuable novel biomarkers in soft tissue sarcomas.

Objective Diffusion weighted imaging (DWI) can be quantified by apparent diffusion coefficient (ADC) and can predict tissue microstructure. The aim of the present study was to analyze possible associations between ADC histogram based parameters with different histopathological parameters in cervical squamous cell carcinoma. Materials and methods 18 female patients (age range 32–79 years) with squamous cell cervical carcinoma were retrospectively enrolled. In all cases, pelvic MRI was performed with a DWI (b-values 0 and 1000 s/mm2). Histogram analysis was performed as a whole lesion measurement. Histopathological parameters included expression of EGFR, VEGF, Hif1-alpha, Her2 and Histone 3. Spearman's correlation coefficient was used to analyze associations between investigated parameters. Results Analyze of the investigated ADC histogram parameters showed a good interreader variability, ranging from 0.705 for entropy to 0.959 for ADCmedian. EGFR expression correlated statistically significant with several histogram parameters. The highest correlation was observed for p75 (p = −0.562, P = 0.015). There were several correlations with histone 3, the highest with p25 (p = −0.610, P = 0.007). None of the ADC related parameters correlated statistically significant with expression of VEGF, Hif1-alpha and Her2. Conclusion Histogram analysis showed a good interreader agreement. ADC histogram parameters might be able to reflect expression of EGFR and histone 3 in cervical squamous cell carcinomas, but not expression of VEGF, Hif1-alpha and Her2.

Perifocal edema of brain tumors is associated with survival and neurological symptoms. Our aim was to analyze associations between perifocal edema and immunohistochemical features including proliferation potential, microvessel density, neoangiogenesis and invasiveness in brain metastasis (BM).35 patients with BM were included into the retrospective study. The tumors were localized supratentorial in 25 lesions (71.4%) and infratentorial in 10 lesions (28.6%). The following immunohistochemical features were calculated on histopathological specimens: microvessel density, proliferation index Ki 67, matrix-metallopeptidase 9 (MMP9) extracellular matrix metalloproteinase inducer (EMMPRIN) and vascular endothelial growth factor (VEGF) expression. Tumor and edema volumes were estimated semiautomatically on magnetic resonance images.There were no correlations between tumor volume and edema volume. Moreover, no correlation was identified between the investigated immunohistochemical features and tumor/edema volume. In the non-small cell lung cancer subgroup, a positive correlation between tumor volume and VEGF expression was observed (r = 0.52, P = 0.02) and edema volume correlated inversely with MMP9 expression (r = -0.53, P = 0.02).In BM, no linear associations exist between tumor volumes, edema volumes and immunohistochemical features reflecting proliferation potential, neoangiogenesis, microvessel density and MMP9 expression. However, in the subgroup of non-small cell lung cancer, there might be associations between MMP9 expression and edema volume as well as between tumor volume and angiogenesis.

Objective Apparent diffusion coefficient (ADC) values derived from Diffusion-weighted images are able to reflect tumor microstructure, such as cellularity, extracellular matrix or proliferation potential. This present study sought to correlate prognostic relevant histopathologic parameters with ADC values derived from a whole lesion measurement in head and neck squamous cell carcinoma (HNSCC). Materials and methods Thirty-four patients with histological proven primary HNSCC were prospectively acquired. Histogram analysis was derived from ADC maps. In all cases, expression of Hif1-alpha, VEGF, EGFR, p53, p16, Her 2 were analyzed. Results In the overall patient sample, ADCmax correlated with p53 expression (p = −0.446, p = 0.009) and ADCmode correlated with Her2-expression (p = −0.354, p = 0.047). In the p16 positive group there were several correlations. P25, P90 and entropy correlated with Hif1-alpha (p = −0.423, p = 0.05, p = −0.494, p = 0.019, p = 0.479, p = 0.024, respectively). Kurtosis correlated with P53 expression (p = −0.466, p = 0.029). For p16 negative carcinomas the following associations could be identified. Mode correlated with VEGF-expression (p = −0.657, p = 0.039). ADCmax, P75, P90, and Std correlated with p53-expression (p = −0.827, p = 0.002, p = −0.736, p = 0.01, p = −0.836, p = 0.001 and p = −0.70, p = 0.016, respectively). There were no statistically significant differences of ADC histogram parameters between p16 positive and p16 negative carcinomas. Conclusion ADC histogram values can reflect different histopathological features in HNSCC. Associations between ADC histogram analysis parameters and histopathology depend on p16 status.

Purpose To analyze several histopathological features and their possible correlations with whole lesion histogram analysis derived from ADC maps in meningioma. Materials and methods The retrospective study involved 36 patients with primary meningiomas. For every tumor, the following histogram analysis parameters of apparent diffusion coefficient (ADC) were calculated: ADCmean, ADCmax, ADCmin, ADCmedian, ADCmode, ADC percentiles: P10, P25, P75, P90, as well kurtosis, skewness, and entropy. All measures were performed by two radiologists. Proliferation index KI 67, minimal, maximal and mean cell count, total nucleic area, and expression of water channel aquaporin 4 (AQP4) were estimated. Spearman's correlation coefficient was used to analyze associations between investigated parameters. Results A perfect interobserver agreement for all ADC values (0.84–0.97) was identified. All ADC values correlated inversely with tumor cellularity with the strongest correlation between P10, P25 and mean cell count (−0.558). KI 67 correlated inversely with all ADC values except ADCmin. ADC parameters did not correlate with total nucleic area. All ADC values correlated statistically significant with expression of AQP4. Conclusions ADC histogram analysis is a valid method with an excellent interobserver agreement. Cellularity parameters and proliferation potential are associated with different ADC values. Membrane permeability may play a greater role for water diffusion than cell count and proliferation activity.

Flow diversion (FD) is a novel endovascular technique based on the profound alteration of cerebrovascular hemodynamics, which emerged as a promising minimally invasive therapy for intracranial aneurysms. However, delayed post-procedural stroke remains an unexplained concern. A consistent follow-up-regimen has not yet been defined, but is required urgently to clarify the underlying cause of delayed ischemia. In the last two years, 223 patients were treated with six different FD devices in our center. We identified subacute, FD-induced segmental vasospasm (SV) in 36 patients as a yet unknown, delayed-type reaction potentially compromising brain perfusion to a critical level. Furthermore, 86% of all patients revealed significant SV approximately four weeks after treatment. In addition, 56% had SV with 25% stenosis, and 80% had additional neointimal hyperplasia. Only 13% exhibited SV-related high-grade stenosis. One of those suffered stroke due to prolonged SV, requiring neurocritical care and repeated intra-arterial (i.a.) biochemical angioplasty for seven days to prevent territorial infarction. Five patients suffered newly manifested, transient hemicrania accompanying a compensatorily increased ipsilateral leptomeningeal perfusion. One treated vessel obliterated permanently. Hence, FD-induced SV is a frequent vascular reaction after FD treatment, potentially causing symptomatic ischemia or even stroke, approximately one month post procedure. A specifically early follow-up-strategy must be applied to identify patients at risk for ischemia, requiring intensified monitoring and potentially anti-vasospastic treatment.

Introduction: Texture analysis is an emergent imaging technique to quantify heterogeneity in radiological images. It is still unclear whether this technique is capable to reflect tumor microstructure. The present study sought to correlate histopathology parameters with texture features derived from contrast-enhanced CT images in head and neck squamous cell carcinomas (HNSCC). Materials and Methods: Twenty-eight patients with histopathological proven HNSCC were retrospectively analyzed. In every case EGFR, VEGF, Hif1-alpha, Ki67, p53 expression derived from immunhistochemical specimen were semiautomatically calculated. Furthermore, mean cell count was estimated. Texture analysis was performed on contrast-enhanced CT images as a whole lesion measurement. Spearman's correlation analysis was performed, adjusted with Benjamini-Hochberg correction for multiple tests. Results: Several texture features correlated with histopathological parameters. After correction only CT texture joint entropy and CT entropy correlation with Hif1-alpha expression remained statistically significant (ρ = -0.60 and ρ = -0.50, respectively). Conclusions: CT texture joint entropy and CT entropy were associated with Hif1-alpha expression in HNSCC and might be able to reflect hypoxic areas in this entity.

Diffusion-weighted imaging (DWI) is a functional imaging technique sensitive to microstructure in tissues. It is widely acknowledged to reflect cellularity in tumors. A small part of DWI is also sensitive to perfusion-related information and might therefore be also be able to reflect microvessel density in tumor tissues. Aim of the present study was to elucidate possible correlations between microvessel density and apparent diffusion coefficient (ADC) values in head and neck squamous cell carcinoma (HNSCC). Thirty-four patients with histologically proven primary HNSCC were included in the study. DWI was performed with a 3 T magnetic resonance imaging (MRI) (b-values 0 and 800 s/mm2) and histogram analysis was calculated with a whole lesion measurement. In every case, microvessel density was estimated with CD105-stained specimens. There were no statistically significant correlations between ADC histogram parameters and microvessel density. The calculated correlation coefficients ranged from r = -0.27, P = .13 for entropy and vessel area to r = 0.16, P = .40 for ADCmin and vessel count. Whole-lesion histogram analysis of ADC values cannot reflect microvessel density in HNSCC.

To analyze associations between histogram analysis parameters derived from conventional magnetic resonance imaging (MRI) and different histopathological features in head and neck squamous cell carcinoma (HNSCC). Thirty-four patients with histologically proven primary HNSCC were prospectively acquired. Histogram analysis was derived from pre-contrast T1-weighted (T1w) and T2-weighted (T2w) images. In all cases, expression of HIF-1α, VEGF, EGFR, p53, Ki67, and p16 as well as tumor cell count was analyzed. In the overall sample, inverse correlation between entropy derived from T1w images and p53 expression (p = − 0.458, P = 0.01) was found. Furthermore, p10 derived from T1w images correlated with VEGF expression (p = 0.371, P = 0.04). In the p16-positive tumors, VEGF expression correlated with several parameters derived from T1w images: mean (p = 0.481, P = 0.032), p10 (p = 0.489, P = 0.029), p25 (p = 0.475, P = 0.034), median (p = 0.468, P = 0.037), and mode (p = 0.492, P = 0.028). Several T2w parameters were associated with p53 expression: mean (p = 0.569, P = 0.007), p25 (p = 0.508, P = 0.019), p75 (p = 0.479, P = 0.028), median (p = 0.555, P = 0.009), and mode (p = 0.468, P = 0.033). Kurtosis derived from T2w images correlated with cell count (p = 0.534, P = 0.013). In p16-negative carcinomas, T2w parameters correlated with p53 expression: max (p = 0.736, P = 0.015), p90 (p = 0.687, P = 0.028), and standard deviation (p = 0.760, P = 0.011). T2w p10 (p = − 0.709, P = 0.022) and T2w p25 (p = − 0.733, P = 0.016) correlated also with HIF-1α expression. Multiple associations between histogram parameters derived from T1w and T2w images and clinically relevant histopathological features were found in HNSCC. Therefore, imaging parameters can be also used as surrogate markers for tumor cellularity, proliferation, and vascularization in HNSCC. The identified correlations differed significantly between p16-positive and p16-negative cancers.

Histogram analysis is an imaging analysis in which a whole tumor can be assessed, and every voxel of a radiological image is issued into a histogram. Thereby, statistically information about tumor can be obtained. The purpose of the study was to analyze possible relationships between histogram parameters derived from conventional MRI sequences and several histopathological features in cervical squamous cell carcinomas.A total of 18 female patients (age range 32-79 years) with squamous cell cervical carcinoma were retrospectively enrolled into the study. In all cases, pelvic MRI with a clinically protocol was performed. Histogram analysis was performed as a whole lesion measurement, calculating several percentils, minimum, mean, median, mode, maximum, kurtosis, skewness, and entropy. Histopathological parameters included expression of epidermal-growth factor (EGFR), vascular endothelial growth factor, hypoxia-inducible factor 1-alpha, Her2, and Histone 3. Spearman's correlation coefficient was used to analyze associations between investigated parameters.Several pre- and postcontrast derived T1-weighted parameters correlated inversely with EGFR expression. For precontrast T1-weighted images, the strongest correlation was found for p90 (ρ = -0.77, p = 0.004). For postcontrast T1-weighted images, the strongest correlation was observed for minimum (ρ = -0.64, p = 0.021). Several parameters derived from T2-weighted images were statistically significant different between Her2-positive and Her2 negative tumors. Skewness had the best p-value ( p = 0.004).Histogram analysis parameters of T1-weighted and T2-weighted images reflect HER2 status and EGFR expression in cervical cancer. Histogram parameters cannot predict cell count, proliferation index, or angiogenesis related histopathological features.

WOS: 000444013200002 PubMed ID: 30005209 Low grade meningiomas have better prognosis than high grade meningiomas. The aim of this study was to measure apparent diffusion coefficient (ADC) histogram analysis parameters in different meningiomas in a large multicenter sample and to analyze the possibility of several parameters for predicting tumor grade and proliferation potential. Overall, 148 meningiomas from 7 institutions were evaluated in this retrospective study. Grade 1 lesions were diagnosed in 101 (68.2%) cases, grade 2 in 41 (27.7%) patients, and grade 3 in 6 (4.1%) patients. All tumors were investigated by MRI (1.5T scanner) by using diffusion weighted imaging (b values of 0 and 1000 s/mm (2)). For every lesion, the following parameters were calculated: mean ADC, maximum ADC, minimum ADC, median ADC, mode ADC, ADC percentiles P10, P25, P75, P90, kurtosis, skewness, and entropy. The comparison of ADC values was performed by Mann Whitney-U test. Correlation between different ADC parameters and KI 67 was calculated by Spearman's rank correlation coefficient. Grade 2/3 meningiomas showed statistically significant lower ADC histogram analysis parameters in comparison to grade 1 tumors, especially ADC median. A threshold value of 0.82 for ADC median to predict tumor grade was estimated (sensitivity = 82.2%, specificity = 63.8%, accuracy = 76.4%, positive and negative predictive values were 83% and 62.5%, respectively). All ADC parameters except maximum ADC showed weak significant correlations with KI 67, especially ADC P25 (P =-.340, P = .0001).

(1) Background: About 15% of the patients undergoing neoadjuvant chemoradiation for locally advanced rectal cancer exhibit pathological complete response (pCR). The surgical approach is associated with major risks as well as a potential negative impact on quality of life and has been questioned in the past. Still, there is no evidence of a reliable clinical or radiological surrogate marker for pCR. This study aims to replicate previously reported response predictions on the basis of non-contrast CT scans on an independent patient cohort. (2) Methods: A total of 169 consecutive patients (126 males, 43 females) that underwent neoadjuvant chemoradiation and consecutive total mesorectal excision were included. The solid tumors were segmented on CT scans acquired on the same scanner for treatment planning. To quantify intratumoral 3D spatial heterogeneity, 1819 radiomics parameters were derived per case. Feature selection and algorithmic modeling were performed to classify pCR vs. non-pCR cases. A random forest model was trained on the dataset using 4-fold cross-validation. (3) Results: The model achieved an accuracy of 87%, higher than previously reported. Correction for the imbalanced distribution of pCR and non-PCR cases (13% and 87% respectively) was applied, yielding a balanced accuracy score of 0.5%. An additional experiment to classify a computer-generated random data sample using the same model led to comparable results. (4) Conclusions: There is no evidence of added value of a radiomics model based on on-contrast CT scans for prediction of pCR in rectal cancer. The imbalance of the target variable could be identified as a key issue, leading to a biased model and optimistic predictions., publishedVersion

Introduction The involvement of the serotonergic system of the brainstem raphe in the pathogenesis of migraine is discussed. Here we studied brainstem alterations in migraineurs using transcranial sonography and examined their relation to clinical features and self-medication. Methods We investigated 51 migraineurs (11 men, 40 women, mean age 29.7 ± 11.9 years) and 32 healthy individuals without history of headache or depressive disorder (eight men, 24 women, mean age 34.4 ± 13.0 years). Transcranial sonography was performed in an investigator-blinded fashion. Midbrain raphe echogenicity was quantified using digitized analysis. Migraine characteristics and the use of analgesics were evaluated by applying validated questionnaires. Eight migraineurs underwent neurophysiologic evaluation of contingent stimulus-related cortical potentials. Results Echo-reduced midbrain raphe was detected in 27 (53%) migraineurs, but only six (19%) control subjects (odds ratio = 4.87, p = 0.002). Lower raphe echogenicity correlated with both higher amplitude of terminal contingent negative variation (Spearman test, r = 0.76, p = 0.028) and higher use of analgesic drugs ( r = −0.45, p = 0.011), but not with use of triptans or with migraine frequency or severity (all p > 0.2). Compared to migraineurs without aura, migraineurs with aura had enlarged third ventricles (t-test, p = 0.014), while the lateral ventricle widths did not differ ( p = 0.62). Conclusions Midbrain raphe alteration is frequent in migraineurs and relates to self-medication behavior. This alteration may reflect the dysfunction of serotonergic raphe nuclei.