Strategies to develop radiomics and machine learning models for lung cancer stage and histology prediction using small data samples

Predictive models based on radiomics and machine-learning (ML) need large and annotated datasets for training, often difficult to collect. We designed an operative pipeline for model training to exploit data already available to the scientific community. The aim of this work was to explore the capability of radiomic features in predicting tumor histology and stage in patients with non-small cell lung cancer (NSCLC). We analyzed the radiotherapy planning thoracic CT scans of a proprietary sample of 47 subjects (L-RT) and integrated this dataset with a publicly available set of 130 patients from the MAASTRO NSCLC collection (Lung1). We implemented intra- and inter-sample cross-validation strategies (CV) for evaluating the ML predictive model performances with not so large datasets. We carried out two classification tasks: histology classification (3 classes) and overall stage classification (two classes: stage I and II). In the first task, the best performance was obtained by a Random Forest classifier, once the analysis has been restricted to stage I and II tumors of the Lung1 and L-RT merged dataset (AUC = 0.72 ± 0.11). For the overall stage classification, the best results were obtained when training on Lung1 and testing of L-RT dataset (AUC = 0.72 ± 0.04 for Random Forest and AUC = 0.84 ± 0.03 for linear-kernel Support Vector Machine). According to the classification task to be accomplished and to the heterogeneity of the available dataset(s), different CV strategies have to be explored and compared to make a robust assessment of the potential of a predictive model based on radiomics and ML.

Lower gray matter volumes of frontal lobes and insula in adolescents with anorexia nervosa restricting type: Findings from a Brain Morphometry Study

BACKGROUND

Brain atrophy in anorexia nervosa (AN) is one of the most marked structural brain changes observed in mental disorders. In this study, we propose a whole brain analysis approach to characterize global and regional cerebral volumes in adolescents with restricting-type anorexia nervosa (AN-r).

METHODS

A total of 48 adolescent females (age range 13-18 years) were enrolled in the study (24 right-handed AN-r in the early stages of the illness and treated in the same clinical setting and 24 age-matched healthy controls [HC]). High-resolution T1-weighted magnetic resonance images were acquired. Cerebral volumes, including the total amounts of gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) were obtained with the Statistical Parametric Mapping software (SPM8); specific cortical regional volumes were computed by applying an atlas-based cortical parcellation to the SPM8 GM segments. Analysis of variance (ANOVA) was performed to identify any significant between-group differences in global and regional brain volumes.

RESULTS

The analyses revealed reduced total GM volumes (p = 0.02) and increased CSF (p = 0.05) in AN-r, compared with HC. No significant between-group difference was found in WM volumes. At the regional level, significantly lower GM volumes in both frontal lobes (p = 0.006) and in the left insula (p = 0.016) were detected. No significant relationships were found between cerebral volumes and duration of illness, psychiatric comorbidities, psychopharmacological treatment, prepubertal phase, or presence of amenorrhea.

CONCLUSIONS

The topographic distribution of GM reduction in a homogenous group of AN-r involves regions responsible for the emotional and cognitive deficits associated with the illness. These findings are discussed in relation to the roles of the insular cortex and the frontal lobes.

Evaluation of the intra- and inter-method agreement of brain MRI segmentation software packages: A comparison between SPM12 and FreeSurfer v6.0

PURPOSE

The lack of inter-method agreement can produce inconsistent results in neuroimaging studies. We evaluated the intra-method repeatability and the inter-method reproducibility of two widely-used automatic segmentation methods for brain MRI: the FreeSurfer (FS) and the Statistical Parametric Mapping (SPM) software packages.

METHODS

We segmented the gray matter (GM), the white matter (WM) and subcortical structures in test-retest MRI data of healthy volunteers from Kirby-21 and OASIS datasets. We used Pearson's correlation (r), Bland-Altman plot and Dice index to study intra-method repeatability and inter-method reproducibility. In order to test whether different processing methods affect the results of a neuroimaging-based group study, we carried out a statistical comparison between male and female volume measures.

RESULTS

A high correlation was found between test-retest volume measures for both SPM (r in the 0.98-0.99 range) and FS (r in the 0.95-0.99 range). A non-null bias between test-retest FS volumes was detected for GM and WM in the OASIS dataset. The inter-method reproducibility analysis measured volume correlation values in the 0.72-0.98 range and the overlap between the segmented structures assessed by the Dice index was in the 0.76-0.83 range. SPM systematically provided significantly greater GM volumes and lower WM and subcortical volumes with respect to FS. In the male vs. female brain volume comparisons, inconsistencies arose for the OASIS dataset, where the gender-related differences appear subtler with respect to the Kirby dataset.

CONCLUSIONS

The inter-method reproducibility should be evaluated before interpreting the results of neuroimaging studies.

Vascular Function Is Improved After an Environmental Enrichment Program

Environmental enrichment may slow cognitive decay possibly acting through an improvement in vascular function. Aim of the study was to assess the effects of a 7-month cognitive, social, and physical training program on cognitive and vascular function in patients with mild cognitive impairment. In a single-center, randomized, parallel-group study, 113 patients (age, 65–89 years) were randomized to multidomain training (n=55) or usual care (n=58). All participants underwent neuropsychological tests and vascular evaluation, including brachial artery flow-mediated dilation, carotid–femoral pulse wave velocity, carotid distensibility, and assessment of circulating hematopoietic CD34+ and endothelial progenitor cells. At study entry, an age-matched control group (n=45) was also studied. Compared with controls, patients had at study entry a reduced flow-mediated dilation (2.97±2.14% versus 3.73±2.06%; P =0.03) and hyperemic stimulus (shear rate area under the curve, 19.1±15.7 versus 25.7±15.1×10 −3 ; P =0.009); only the latter remained significant after adjustment for confounders ( P =0.03). Training improved Alzheimer disease assessment scale cognitive (training, 14.0±4.8 to 13.1±5.5; nontraining, 12.1±3.9 to 13.2±4.8; P for interaction visit×training=0.02), flow-mediated dilation (2.82±2.19% to 3.40±1.81%, 3.05±2.08% to 2.24±1.59%; P =0.006; P =0.023 after adjustment for diameter and shear rate area under the curve), and circulating hematopoietic CD34 + cells and prevented the decline in carotid distensibility (18.4±5.3 to 20.0±6.6, 23.9±11.0 to 19.5±7.1 Pa −1 ; P =0.005). The only clinical predictor of improvement of cognitive function after training was established hypertension. There was no correlation between changes in measures of cognitive and vascular function. In conclusion, a multidomain training program slows cognitive decline, especially in hypertensive individuals. This effect is accompanied by improved systemic endothelial function, mobilization of progenitor CD34 + cells, and preserved carotid distensibility.

Clinical Trial Registration—

URL: http://www.clinicaltrials.gov . Unique identifier: NCT01725178.

GPCALMA: a Grid-based Tool for Mammographic Screening

Objectives: The next generation of high energy physics (HEP) experiments requires a GRID approach to a distributed computing system: the key concept is the Virtual Organisation (VO), a group of distributed users with a common goal and the will to share their resources.

Methods: A similar approach, applied to a group of hospitals that joined the GPCALMA project (Grid Platform for Computer Assisted Library for MAmmography), will allow common screening programs for early diagnosis of breast and, in the future, lung cancer. The application code makes use of neural networks for the image analysis and is useful in improving the radiologists' diagnostic performance. GRID services allow remote image analysis and interactive online diagnosis, with a potential for a relevant reduction of the delays presently associated with screening programs.

Results and Conclusions: A prototype of the system, based on AliEn GRID Services [1], is already available, with a central server running common services [2] and several clients connecting to it. Mammograms can be acquired in any location; the related informatio required to select and access them at any time is stored in a common service called Data Catalogue, which can be queried by any client. Thanks to the PROOF facility [3], the result of a query can be used as input for analy-sis algorithms, which are executed on the nodes where the input images are stored,. The selected approach avoids data transfers for all the images with a negative diagnosis and allows an almost real time diagnosis for the set of images with high cancer probability.

Semiautomated Evaluation of the Primary Motor Cortex in Patients with Amyotrophic Lateral Sclerosis at 3T

BACKGROUND AND PURPOSE

Amyotrophic lateral sclerosis is a neurodegenerative disease involving the upper and lower motor neurons. In amyotrophic lateral sclerosis, pathologic changes in the primary motor cortex include Betz cell depletion and the presence of reactive iron-loaded microglia, detectable on 7T MR images as atrophy and T2*-hypointensity. Our purposes were the following: 1) to investigate the signal hypointensity-to-thickness ratio of the primary motor cortex as a radiologic marker of upper motor neuron involvement in amyotrophic lateral sclerosis with a semiautomated method at 3T, 2) to compare 3T and 7T results, and 3) to evaluate whether semiautomated measurement outperforms visual image assessment.

MATERIALS AND METHODS

We investigated 27 patients and 13 healthy subjects at 3T, and 19 patients and 18 healthy subjects at 7T, performing a high-resolution 3D multiecho T2*-weighted sequence targeting the primary motor cortex. The signal hypointensity-to-thickness ratio of the primary motor cortex was calculated with a semiautomated method depicting signal intensity profiles of the cortex. Images were also visually classified as "pathologic" or "nonpathologic" based on the primary motor cortex signal intensity and thickness.

RESULTS

The signal hypointensity-to-thickness ratio of the primary motor cortex was greater in patients than in controls (P < .001), and it correlated with upper motor neuron impairment in patients (ρ = 0.57, P < .001). The diagnostic accuracy of the signal hypointensity-to-thickness ratio was high at 3T (area under the curve = 0.89) and even higher at 7T (area under the curve = 0.94). The sensitivity of the semiautomated method (0.81) outperformed the sensitivity of the visual assessment (0.56-0.63) at 3T.

CONCLUSIONS

The signal hypointensity-to-thickness ratio of the primary motor cortex calculated with a semiautomated method is suggested as a radiologic marker of upper motor neuron burden in patients with amyotrophic lateral sclerosis. This semiautomated method may be useful for improving the subjective radiologic evaluation of upper motor neuron pathology in patients suspected of having amyotrophic lateral sclerosis.

A completely automated CAD system for mass detection in a large mammographic database

Mass localization plays a crucial role in computer-aided detection (CAD) systems for the classification of suspicious regions in mammograms. In this article we present a completely automated classification system for the detection of masses in digitized mammographic images. The tool system we discuss consists in three processing levels: (a) Image segmentation for the localization of regions of interest (ROIs). This step relies on an iterative dynamical threshold algorithm able to select iso-intensity closed contours around gray level maxima of the mammogram. (b) ROI characterization by means of textural features computed from the gray tone spatial dependence matrix (GTSDM), containing second-order spatial statistics information on the pixel gray level intensity. As the images under study were recorded in different centers and with different machine settings, eight GTSDM features were selected so as to be invariant under monotonic transformation. In this way, the images do not need to be normalized, as the adopted features depend on the texture only, rather than on the gray tone levels, too. (c) ROI classification by means of a neural network, with supervision provided by the radiologist's diagnosis. The CAD system was evaluated on a large database of 3369 mammographic images [2307 negative, 1062 pathological (or positive), containing at least one confirmed mass, as diagnosed by an expert radiologist]. To assess the performance of the system, receiver operating characteristic (ROC) and free-response ROC analysis were employed. The area under the ROC curve was found to be Az = 0.783 +/- 0.008 for the ROI-based classification. When evaluating the accuracy of the CAD against the radiologist-drawn boundaries, 4.23 false positives per image are found at 80% of mass sensitivity.

GPCALMA: a Grid-based tool for mammographic screening

OBJECTIVES

The next generation of high energy physics (HEP) experiments requires a GRID approach to a distributed computing system: the key concept is the Virtual ORGANISATION (VO), a group of distributed users with a common goal and the will to share their resources.

METHODS

A similar approach, applied to a group of hospitals that joined the GPCALMA project (Grid Platform for Computer Assisted Library for MAmmography), will allow common screening programs for early diagnosis of breast and, in the future, lung cancer. The application code makes use of neural networks for the image analysis and is useful in improving the radiologists' diagnostic performance. GRID services allow remote image analysis and interactive online diagnosis, with a potential for a relevant reduction of the delays presently associated with screening programs.

RESULTS AND CONCLUSIONS

A prototype of the system, based on AliEn GRID Services [1], is already available, with a central server running common services [2] and several clients connecting to it. Mammograms can be acquired in any location; the related information required to select and access them at any time is stored in a common service called Data Catalogue, which can be queried by any client. Thanks to the PROOF facility [3], the result of a query can be used as input for analysis algorithms, which are executed on the nodes where the input images are stored,. The selected approach avoids data transfers for all the images with a negative diagnosis and allows an almost real time diagnosis for the set of images with high cancer probability.

Antifungal agents. 9. 3-Aryl-4-[alpha-(1H-imidazol-1-yl)arylmethyl]pyrroles: a new class of potent anti-Candida agents

A new class of potent antifungal agents, namely, 3-aryl-4-[alpha-(1H-imidazol-1-yl)arylmethyl]-pyrroles, is described. These compounds are related to bifonazole and pyrrolnitrin, two compounds belonging to the class of antimycotic drugs. The synthesis of the title pyrroles has been performed starting from 1,3-diaryl-2-propen-1-ones, which were reacted with tosylmethyl isocyanide to give 3-aroyl-4-arylpyrroles. Reduction of the resulting compounds by lithium aluminum hydride furnished the related alcohols, which were treated with 1,1'-carbonyldimidazole to afford the required imidazole derivatives. Forty-four new pyrroles which incorporate an (arylmethyl)imidazole moiety in the 3-arylpyrrole structure were prepared by the above procedure and tested in vitro against Candida albicans and Candida spp. Among test compounds, 10 were found to be highly active against C. albicans. The most active derivative (27) was twice as potent (MIC90) as bifonazole, and its activity was 4 times greater than those of miconazole and ketoconazole. The other nine compounds showed antifungal activity of the same order of that of bifonazole and were ca. 2 times as active as miconazole and ketoconazole. Derivatives 21 and 27 tested in vivo against C. albicans A170 were shown to be highly effective in rabbit skin candidosis. Pharmacological studies on compounds 27 and other related pyrroles (19, 35, 36, 38, 39, and 49) are in progress to select one of them as a potential candidate for clinical experiments.

Antifungal agents. VIII. Synthesis and antifungal activities of bipyrryl analogues of bifonazole

Various bipyrryl analogues of bifonazole were synthesized starting from aryl-3-pyrryl-1-imidazolylmethanes. The introduction of a second pyrryl portion was performed by linking an acrylate moiety at 1-position of the pyrrole ring and then by treatment with TosMIC. The bipyrryl esters were hydrolyzed and decarboxylated to afford the required imidazoles. All new imidazole derivatives were tested against Candida albicans and Candida spp using as standard controls miconazole, bifonazole and ketoconazole.

Antifungal agents. 7. Dichlorophenylpyrrylimidazolylmethane derivatives: synthesis and antifungal activities

The synthesis and antifungal activities of some dichlorophenyl-1H-pyrrol-2-yl-1H-imidazol-1-ylmethane derivatives substituted at pyrrole nitrogen are reported. When tested against Candida albicans and Candida spp., some derivatives were found to be from two to four times less active than miconazole, bifonazole and ketoconazole, used as standard controls.

Antifungal agents, II: Synthesis and antifungal activities of aryl-1H-pyrrol-2-yl-1H-imidazol-1-yl-methane derivatives with unsaturated chains

The synthesis and antifungal activities of aryl-1H-pyrrol-2-yl-1H-imidazol-1-yl-methanes having allyl, crotyl, and acrylate chains linked to the N-pyrrole atom and substituted at phenyl ring by Cl, F, CH3, and NO2 groups are reported. In vitro tests against Candida albicans and Candida spp. showed 2,4-dichlorophenyl-1-allyl-1H-pyrrol-2-yl-1H-imidazol-1-yl-methane to be the most potent derivative with activities comparable to those of ketoconazole and slightly inferior to those of bifonazole and miconazole. Some structure-activity relationships are discussed.

Antifungal agents. VI. In vitro antifungal activities of halobenzoyl esters of cis- and trans-[2-(1,1'-biphenyl-4-yl)-2-(1H-imidazol-1-ylmethyl)-1,3- dioxolan-4-yl]carbinols

The synthesis and the in vitro antifungal activities against Candida albicans and Candida spp of a number of halobenzoyl esters of cis- and trans- [2-(1,1'-biphenyl-4-yl)-2-(1H-imidazol-1-ylmethyl)-1,3-dioxolan -4- yl]carbinols is reported. Some new imidazoles were found more active than ketoconazole and sometimes as potent as bifonazole against Candida albicans. All derivatives were found scarcely active against Candida spp.

Antifungal agents. III. Naphthyl and thienyl derivatives of 1H-imidazol-1-yl-4-phenyl-1H-pyrrol-3-ylmethane

The in vitro antifungal activities of some naphthyl and thienyl derivatives of 1H-imidazol-1-yl-4-phenyl-1H-pyrrol-3-ylmethane against Candida albicans and Candida spp are reported. The title derivatives were prepared starting from proper arylstirylketones, which were reacted with tosylmethylisocyanide (Tos-MIC) to afford the related 4-phenyl-1H-pyrrol-3-yl aryl ketones. Reduction of ketones to the corresponding carbinols followed by treatment of the last compounds with 1,1'-carbonyldiimidazole (CDI) gave the title imidazoles. The related N-methylpyrrole derivatives are also described.

Researches on antibacterial and antifungal agents, XIV: Thiophene analogues of bifonazole

Some thiophene analogues of bifonazole have been synthesized by standard procedures and their antifungal activity has been tested against Candida albicans. Among test derivatives biphenyl-4-yl-5-chloro-thien-2-ylimidazol-1-ylmethane and its 5-deschlorothien-2-yl analogue resulted to be the most active. Their antifungal potency was almost comparable to that of control substances, such as miconazole, ketoconazole, and bifonazole. Replacement of benzene by the pyrrole ring in the biphenyl portion retained almost quantitatively the antifungal activity, whereas substitution with other azoles and with nitrogen alicyclic rings led always to less potent derivatives.

Macrophage lysozyme stimulation by fructose-1-6-diphosphate

FDP produces an increase of serum lysozyme concentration which may be related to stimulation of the phagocytic activity. Mice macrophages in vitro produce extracellular and intracellular LSZ (lysozyme) and FDP (fructose-1-6-diphosphate) increases this production. Also in vivo FDP stimulates the macrophages intracellular lysozyme production. The toxic activity in vitro and the protection in vivo against Staphylococcus pyogenes after FDP administration can also be related to macrophage stimulation.

Action of the ketoconazole-alkylaminoethylglycine combination on Candida albicans

High concentrations of ketoconazole (Ktc) exhibit poor antimicrobial activity in microbicidal test. Alkylaminoethylglycine (AAEG), an amphoteric surface-active agent, employed in subinhibitory concentrations, can enhance the action of ketoconazole. The Ktc-AAEG combination causes a higher rate of potassium ion release compared to both drugs separately. The increased ketoconazole activity may be ascribed to the membrane damage indicated by increased permeability to potassium ions.

[Effect of propyl gallate on the antibacterial activity of meclocycline sulfosalicylate]

Propyl gallate shows little antibacterial activity however it markedly potentiates the activity of meclocycline against Pseudomonas, Proteus, Serratia, Escherichia coli and Klebsiella strains. The potentiating effect of propyl gallate is seen especially with resistant strains whereas sensitive strains of Salmonella do not manifest a potentiating effect on meclocycline by propyl gallate. The mechanism of action of propyl gallate is discussed.