1,721,073 research outputs found
RADIOMICS - Local RADiomic features extracted from multimodal sequences to characterIze heterOgeneity of tuMour habItat and produCe imaging biomarkerS
No full textImaging diagnostic has entered a new era with the availability of high-performance GPUs on entry level workstations, which has permitted to extract a huge number of information from medical images often using well-established machine learning techniques. Our research group (CVG) has been working for more than 20 years in multi-feature analyses of medical and biomedical images, in the research field newly named radiomics, to develop diagnostic, prognostic and predictive imaging biomarkers in breast, lung, liver, prostate, pancreatic, oesophageal and gastroesophageal cancer. In fact, image phenotyping can reveal information regarding the underlying tissue's texture, tumour physiology, or even molecular properties of cancer that are concealed from visual inspection. The RADIOMICS Project aims at exploiting our well-established quantitative imaging, machine learning and artificial intelligence techniques to investigate local radiomic heterogeneity of tumour habitat using single- and multi-modal imaging (e.g. MR, mpMR, CT, CT perfusion, PET) and hybrid technologies (CT/PET, PET/MR). The goal is developing diagnostic, prognostic and predictive imaging biomarkers, also providing a relevant contribution for patient's stratification and staging.Imaging diagnostic has entered a new era with the availability of high-performance GPUs on entry level workstations, which has permitted to extract a huge number of information from medical images often using well-established machine learning techniques. Our research group (CVG) has been working for more than 20 years in multi-feature analyses of medical and biomedical images, in the research field newly named radiomics, to develop diagnostic, prognostic and predictive imaging biomarkers in breast, lung, liver, prostate, pancreatic, oesophageal and gastroesophageal cancer. In fact, image phenotyping can reveal information regarding the underlying tissue's texture, tumour physiology, or even molecular properties of cancer that are concealed from visual inspection.
The RADIOMICS Project aims at exploiting our well-established quantitative imaging, machine learning and artificial intelligence techniques to investigate local radiomic heterogeneity of tumour habitat using single- and multi-modal imaging (e.g. MR, mpMR, CT, CT perfusion, PET) and hybrid technologies (CT/PET, PET/MR). The goal is developing diagnostic, prognostic and predictive imaging biomarkers, also providing a relevant contribution for patient's stratification and staging
PHENOMICS - Cancer stress PHENomics: autOmatic Microscopy Image analysis for in vitro phenotypization of heterogeneity of stressed tumour CellS
No full textPreclinical in vitro studies typically assay cell viability and proliferation, cell cycle perturbation, programmed cell death triggering, and gene expression modulation of cell samples exposed to physical or chemical treatments. In recent years, the emerging discipline of tissue phenomics has stimulated clinical research, improved clinical drug development and treatment of patients, with a great potential impact on oncology. Analogously, single-cell phenomics by microscope image analysis has the potential to positively impact on in vitro studies, offering a quantitative evaluation of treatment effects. In particular, cell imaging opens to unprecedented spatial phenomization, to capture tumour response heterogeneities at subcellular level, by selecting proper phenomic features.
Anticancer treatments share the common goal to selectively stress, and ultimately damage, cancer cells, while not perturbing the physiological state of healthy ones. The aim of PHENOMICS is to characterize cancer cell response to different stress conditions by microscope image analysis, in order to quantify stress-induced alterations in the distribution of phenomic markers. Image-based phenomization will increase the benefit/cost ratio of the studies, by providing more robust and consistent data while reducing times and costs, ultimately boosting the research of effective treatments and speeding up translation of results into next preclinical and clinical trial phases
Extracting equilibrium Blood Flow (eBF) values of CTp liver perfusion from tissue peak: a new promising perfusion parameter
No full textAims and objectives: To evaluate the potentiality of the new BF-based feature we computed, the equilibrium Blood Flow (eBF), which measures BF values at the arterial inflow-venous output balance, to improve reliability of liver CT perfusion values in a cohort of patients with colorectal cancer. Methods and materials: 46 patients with colorectal cancer underwent axial CTp examinations of normal liver at diagnosis. A voxel-based and patient-oriented non parametric model of the tissue time concentration curves (TCCs) are obtained by an in-house algorithm directly exploiting the native CTp Hounsfield Unit (HU) first-pass data. The tissue eBF values are achieved with normalizing the TCCs peak by the area under the TCC. For comparison purposes, the “common” BF is computed using the maximum slope. Overall mean and standard deviation are calculated, the coefficient of variation (CV) is computed to assess means repeatability. Correlations between BF and eBF colormaps are evaluated through Pearson (r). Results: eBF computed on the whole cohort of 46 patients yields CV=13%, almost half than MS, where CV=21%. Mean values for eBF[119±16]ml/min/100g are lower than BF[129±27]ml/min/100g, this being expected, because eBF include outflow measurements. Nonetheless, eBF and BF colormaps show excellent correlations (r>0.97) for 42 patients, and 0.80≤r<0.90 in 4 cases only. Conclusion: eBF allows for the vascular outflow measurements, as confirmed by mean values lower than BF ones, although their colormaps are highly correlated. eBF measures are also much more repeatable (almost twice) than BF ones, this making eBF a new promising perfusion feature that could benefit all CTp studies and, ultimately, CTp standardization
A patient-driven approach to improve reproducibility of blood flow values in CT perfusion of liver
No full textAims and objectives: Replacing model fitting with patient data to improve reproducibility of Blood Flow (BF) values computed with different software in Computed Tomography perfusion (CTp) studies of liver. Methods and materials: 46 patients with colorectal cancer and normal liver underwent a CTp examination at diagnosis, consisting in 60 scans lasting 120 sec (1scan/sec for the first 30 sec, 1scan/3sec after). A first passage analysis is considered. While proper parametric models are used to represent the dynamic (i.e., Time Concentration curves, TCCs) of Contrast Agent (CA) in arterial and portal input, the tissue TCCs were achieved by convolving the voxel-based response function (IRF) characteristic of the patient’s liver with portal and arterial models. In-house deconvolution (DV) and maximum slope (MS) software was used to compute BF values. ANOVA was employed to measure MS and DV reproducibility (p-value>0.05) with this method against classical model fitting. Linear correlation is also measured (R2) between voxel-based MS and DV BF colormaps. Results: ANOVA reports reproducibility of mean BF computed with MS and DV for of both methods, with p-value=0.85 (patient-based) and p-value=0.24 (classical), with excellent correlation (R2≥0.99) in 42 and 28 examinations, respectively. Conclusion: Exploiting the original patient's enhancement curves rather than their mathematical representation allows robust perfusion software to improve voxel-based reproducibility of BF colormaps of single patients as well as the reproducibility of the mean BF in the whole cohort. These findings make CTp to take a step forward towards standardization and precision medicine as well
Analysis of the effects of fitting errors of DCE-CT signals on perfusion parameters
No full textComputed Tomography perfusion (CTp) is a promising technique for estimating perfusion parameters, by analysing Time Concentration Curves (TCCs) of the administered contrast agent. However, several artefacts can degrade the signal quality, jeopardizing quantitative measurements. Despite different methods exploit TCCs to compute perfusion parameters, none of them has investigated how TCC fitting errors may affect final perfusion values. The first goal of this work is to investigate residuals distributions in significant signal’s portions, then relating them to Blood Flow (BF). The Gamma Variate (GV) function is addressed to fit TCCs. Voxel-based BF is computed with the two most spread methods in literature, Maximum Slope (MS) and Deconvolution (DV). Experimental results prove that residuals coming from a Gaussian distribution yield percent errors maps locally smooth, thus attaining residuals-independent BF values. Besides results, the methodological approach can be spent in future researches
in order to encourage CTp reproducibility
A novel algorithm to detect the baseline value of a time signal in Dynamic Contrast Enhanced-Computed Tomography
No full textDynamic Contrast Enhanced-Computed Tomography (DCE-CT) is a functional imaging technique that has aroused a great interest in several clinical applications. The unenhanced portion of DCE-CT signal, the baseline, plays a fundamental role for signal analysis as well as to achieve accurate clinical parameter values, such as perfusion ones, used for diagnosis and prognosis purposes. In this study, a new adaptive iterative algorithm to compute voxel-based baseline values exploiting the maximum number of samples, adaptively for each voxel, is proposed and compared against the three main approaches used in the literature, over a dataset of 30 DCE-CT perfusion (briefly, CTp) liver examinations. Results were evaluated according to classical statistical indexes and tests. The experiments show that voxel-based results achieved by applying the four approaches significantly differ and the error indexes related to our method
are the lowest ones. Our results would expectedly improve the accuracy of all methods, including CTp, relying on the whole signal for computation of clinical parameters
Optimizing parameters of a motion detection system by means of a genetic algorithm
No full textVisual surveillance and monitoring have aroused interest in the computer video community for many years. The
main task of these applications is to identify (and track) moving targets. The traffic monitoring application we
have developed requires that a large
number of parameters is tuned in order to work properly. About thirty
parameters concerning the detection algorithm have been c
onsidered as to be optimized. Accordingly, this paper
shows how a
Genetic Algorithm (GA)
represents a powerful task in order to automatically compute sub-optimal
parameter settings in a motion detection system. Besides, to
our knowledge this work is the first attempt of using
GAs to such a problem. Accurate experiments accomplis
hed on a challenging test se
quence show the relevant
results attained in terms of qualitative performance
Reproducibility of Computed Tomography perfusion parameters in hepatic multicentre study in patients with colorectal cancer
Full text linkObjective: The Computed Tomography perfusion (CTp) is a promising tool in oncology to characterize tissue hemodynamics, but the difficulty to achieve reproducible perfusion parameters in several organs, with different methods, contributes to hamper the clinical translation of CTp. The goal of this study is to set up a new approach aiming at achieving multicentre reproducibility of blood flow (BF) values in liver.
Methods: 75 patients from two Centres (A and B) underwent an axial liver CTp, including arterial and portal phases. A dedicated workflow addressing modelling and computational aspects was implemented, including a novel two-stage strategy to separate the dual-input contributions of hepatic signals, thus allowing to compute independently both Maximum Slope (MS) and Deconvolution (DV) on the same contributing signals.
Results: 95% of patients in A and B showed an excellent voxel-based Pearson correlation ( ≥ 0.96) between MS and DV BF values, with very low coefficients of variation ( = 0.11 in the worst case). The good concordance is confirmed for the whole cohorts, in single Centres and both, where 2=0.97, ≥ 0.97, ≥ 0.96, ≥ 0.78 and =0.25 are the worst values. Compared with eighteen recent articles, these represent by far the best outcomes.
Conclusion: The excellent patient- and cohort-based reproducibility of BF values achieved independently by MS and BV confirms the effectiveness of the approach presented.
Significance: Our approach can be used to improve the reproducibility in other CTp multicentre studies, in liver as well as in other organs, with even different clinical questions, and represents a marked step forward towards CTp standardization, favouring the investigation of imaging biomarkers
Colour vignetting correction for microscopy image mosaics used for quantitative analyses
Full text linkImage mosaicing permits to achieve one high resolution image, extending the visible area of the sample while keeping the same resolution. However, intensity inhomogeneity of the stitched images can alter measurements and the right perception of the original sample. The problem can be solved by flat-field correcting the images through the vignetting function. Vignetting correction has been widely addressed for grey-level images, but not for colour ones. In this work, a practical solution for the colour vignetting correction in microscopy, also facing the problem of saturated pixels, is described. In order to assess the quality of the proposed approach, five different tonal correction approaches were quantitatively compared using state-of-the-art metrics and seven pairs of partially overlapping images of seven different samples. The results obtained proved that the proposed approach allows obtaining high quality colour flat-field corrected images and seamless mosaics without employing any blending adjustment. In order to give the opportunity to easily obtain seamless mosaics ready for quantitative analysis, the described vignetting correction method has been implemented in an upgraded release of MicroMos (version 3.0, http://sourceforge.net/p/micromos), an open-source software specifically designed to automatically obtain mosaics of partially overlapped images
Colormaps of CT perfusion parameters computed using different methods visually match
No full textThe Computed Tomography perfusion (CTp) is a promising tool in oncology to characterize hemodynamics of tissues, based on fast and repeated CT scans of the region of interest after contrast agent administration. However, it has difficulty in entering the clinical routine (substantially, except for brain and heart) because of the difficulty of achieving same perfusion maps as equipment or perfusion software change. In this work, we present a proper computing chain for CTp parameters, relying on a robust and accurate voxel-based computation, that permits two widely used techniques, Maximum
Slope (MS) and Deconvolution (DV) to reproduce, for the first time, the same perfusion maps. The experiments carried out on 25 examinations of oncologic patients proved an excellent correlation between MS and DV maps, with the worst R^2 = 0:971. This outcome represents a marked step forward
in the standardization of CTp studies and encourages further multi-centre analyses
- …
