Using rADioMIcs and machine learning with ultrasonography for the differential diagnosis of myometRiAL tumors (the ADMIRAL pilot study). Radiomics and differential diagnosis of myometrial tumors.

OBJECTIVE: To develop and evaluate the performance of a radiomics and machine learning model applied to ultrasound (US) images in predicting the risk of malignancy of a uterine mesenchymal lesion. METHODS: Single-center retrospective evaluation of consecutive patients who underwent surgery for a malignant uterine mesenchymal lesion (sarcoma) and a control group of patients operated on for a benign uterine mesenchymal lesion (myoma). Radiomics was applied to US preoperative images according to the International Biomarker Standardization Initiative guidelines to create, validate and test a classification model for the differential diagnosis of myometrial tumors. The TRACE4 radiomic platform was used thus obtaining a full-automatic radiomic workflow. Definitive histology was considered as gold standard. Accuracy, sensitivity, specificity, AUC and standard deviation of the created classification model were defined. RESULTS: A total of 70 women with uterine mesenchymal lesions were recruited (20 with histological diagnosis of sarcoma and 50 myomas). Three hundred and nineteen radiomics IBSI-compliant features were extracted and 308 radiomics features were found stable. Different machine learning classifiers were created and the best classification system showed Accuracy 0.85 ± 0.01, Sensitivity 0.80 ± 0.01, Specificity 0.87 ± 0.01, AUC 0.86 ± 0.03. CONCLUSIONS: Radiomics applied to US images shows a great potential in differential diagnosis of mesenchymal tumors, thus representing an interesting decision support tool for the gynecologist oncologist in an area often characterized by uncertainty.

Prognostic Value of 18F-Fluorocholine PET Parameters in Metastatic Castrate-Resistant Prostate Cancer Patients Treated with Docetaxel.

Background and Aim: The availability of new treatments for metastatic castrate-resistant prostate cancer (mCRPC) patients increases the need for reliable biomarkers to help clinicians to choose the better sequence strategy. The aim of the present retrospective and observational work is to investigate the prognostic value of 18F-fluorocholine (18F-FCH) positron emission tomography (PET) parameters in mCRPC. Materials and Methods: Between March 2013 and August 2016, 29 patients with mCRPC were included. They all received three-weekly docetaxel after androgen deprivation therapy, and they underwent 18F-FCH PET/computed tomography (CT) before and after the therapy. Semi-quantitative indices such as maximum standardized uptake value (SUVmax), mean standardized uptake value (SUVmean) with partial volume effect (PVC-SUV) correction, metabolically active tumour volume (MATV), and total lesion activity (TLA) with partial volume effect (PVC-TLA) correction were measured both in pre-treatment and post-treatment 18F-FCH PET/CT scans for each lesion. Whole-body indices were calculated as sum of values measured for each lesion (SSUVmax, SPVC-SUV, SMATV, and STLA). Progression-free survival (PFS) and overall survival (OS) were considered as clinical endpoints. Univariate and multivariate hazard ratios for whole-body 18F-FCH PET indices were performed, and p < 0.05 was considered as significant. Results: Cox regression analysis showed a statistically significant correlation between PFS, SMATV, and STLA. No correlations between OS and 18F-FCH PET parameters were defined probably due to the small sample size. Conclusions: Semi-quantitative indices such as SMATV and STLA at baseline have a prognostic role in patients treated with docetaxel for mCRPC, suggesting a potential role of 18F-FCH PET/CT imaging in clinical decision-making.

A wrapped multi-label classifier for the automatic diagnosis and prognosis of Alzheimer's disease.

BACKGROUND: AD is the most frequent neurodegenerative disease, severely impacting our society. Early diagnosis and prognosis are challenging tasks in the management of AD patients. NEW METHOD: We implemented a machine-learning classifier for the automatic early diagnosis and prognosis of AD by means of features extracted, selected and optimized from structural MRI brain images. The classifier was designed to perform multi-label automatic classification into the following four classes: HC, ncMCI, cMCI, and AD. RESULTS: From our analyses, it emerged that MMSE and hippocampus-related measures must be included as primary measures in automatic-classification systems for both the early diagnosis and the prognosis of AD. The voting scheme mainly based on the binary-classification performances on the different four groups is the best choice to model the multi-label decision function for AD, when compared with a simple majority-vote scheme or with a scheme aimed at discriminating patients with high vs low risk of conversion to AD and therapy addressing. COMPARISON WITH EXISTING METHOD(S): The accuracies of our binary classifications were higher than or comparable to previously published methods. An improvement is needed on the approach we used to combine binary-classification outputs to obtain the final multi-label classification. CONCLUSIONS: The performance of multi-label automatic-classification systems strongly depends on the choice of the voting scheme used for combining binary-classification labels.

An Adaptive Thresholding Method for BTV Estimation Incorporating PET Reconstruction Parameters: A Multicenter Study of the Robustness and the Reliability.

OBJECTIVE: The aim of this work was to assess robustness and reliability of an adaptive thresholding algorithm for the biological target volume estimation incorporating reconstruction parameters. METHOD: In a multicenter study, a phantom with spheres of different diameters (6.5-57.4 mm) was filled with (18)F-FDG at different target-to-background ratios (TBR: 2.5-70) and scanned for different acquisition periods (2-5 min). Image reconstruction algorithms were used varying number of iterations and postreconstruction transaxial smoothing. Optimal thresholds (TS) for volume estimation were determined as percentage of the maximum intensity in the cross section area of the spheres. Multiple regression techniques were used to identify relevant predictors of TS. RESULTS: The goodness of the model fit was high (R(2): 0.74-0.92). TBR was the most significant predictor of TS. For all scanners, except the Gemini scanners, FWHM was an independent predictor of TS. Significant differences were observed between scanners of different models, but not between different scanners of the same model. The shrinkage on cross validation was small and indicative of excellent reliability of model estimation. CONCLUSIONS: Incorporation of postreconstruction filtering FWHM in an adaptive thresholding algorithm for the BTV estimation allows obtaining a robust and reliable method to be applied to a variety of different scanners, without scanner-specific individual calibration.

Partial volume corrected 18F-FDG PET mean standardized uptake value correlates with prognostic factors in breast cancer.

AIM: The aim of this paper was to assess the prognostic role of pretherapy partial volume corrected (PVC) 18F-fluorodeoxyglucose mean standardized uptake value (SUV) in breast cancer (BC). METHODS: Forty oncological patients, BC diagnosed by biopsy, with breast tumor mass diameter >1 cm measured to the mammography, designed for surgical intervention, underwent a pretherapy semi-quantitative 18F-FDG positron emission tomography/computed tomography (18F-FDG PET/CT) whole-body study for tumor staging. Mean Body-Weight Standardized Uptake Value with Correction for Partial Volume effect (PVC- SUVBW-mean) was calculated in all mammary detected lesions. Excised tissues from primitive BC were sectioned and classified according to the WHO guidelines, evaluating biological features. Univariate (Mann-Withney/Kruskal-Wallis) and multivariate (linear regression, hierarchical clustering) statistical tests were performed between PVC-SUVBW-mean and biological indexes. ROC analysis was performed. PVC-SUVBW-mean thresholds were derived allowing to distinguish groups of BC patients with different biological characteristics. Specificity and Sensitivity were also calculated. RESULTS: Statistical and multiple correlations between pretherapy 18F-FDG PET PVC-SUVBW-mean and histological type, grade, ER/PgR hormone receptors and Mib-1 cellular proliferation index were found. In our samples, PVC-SUVBW-mean <≈4 g/cc was found correlated to BC patients with Invasive Lobular Carcinoma (ILC) or well differentiated Invasive Ductal Carcinoma (IDC), a positive expression of ER and PgR and a negative expression of MiB-1, while PVC-SUVBW-mean >≈7.00 is associated to BC patients with moderately and poorly differentiated IDC, negative expression of ER and PgR and a positive expression of MiB-1. CONCLUSION: Pretherapy PVC 18F-FDG PET PVC-SUVBW-mean measurement correlates with prognostic factors in BC and could be used to stratify patients before intervention.

Predictive value of pre-therapy (18)F-FDG PET/CT for the outcome of (18)F-FDG PET-guided radiotherapy in patients with head and neck cancer.

PURPOSE: The aim of this study was to evaluate the predictive role of pre-therapy fluorodeoxyglucose (FDG) uptake parameters of primary tumour in head and neck cancer (HNC) patients undergoing intensity-modulated radiotherapy (IMRT) with simultaneous integrated boost (SIB) on FDG-positive volume-positron emission tomography (PET) gross tumour volume (PET-GTV). METHODS: This retrospective study included 19 patients (15 men and 4 women, mean age 59.2 years, range 23-81 years) diagnosed with HNC between 2005 and 2011. Of 19 patients, 15 (79 %) had stage III-IV. All patients underwent FDG PET/CT before treatment. Metabolic indexes of primary tumour, including metabolic tumour volume (MTV), maximum and mean standardized uptake value (SUVmax, SUVmean) and total lesion glycolysis (TLG) were considered. Partial volume effect correction (PVC) was performed for SUVmean and TLG estimation. Correlations between PET/CT parameters and 2-year disease-free survival (DFS), local recurrence-free survival (LRFS) and distant metastasis-free survival (DMFS) were assessed. Median patient follow-up was 19.2 months (range 4-24 months). RESULTS: MTV, TLG and PVC-TLG predicting patients' outcome with respect to all the considered local and distant disease control endpoints (LRFS, DMFS and DFS) were 32.4 cc, 469.8 g and 547.3 g, respectively. SUVmean and PVC-SUVmean cut-off values predictive of LRFS and DFS were 10.8 and 13.3, respectively. PVC was able to compensate errors up to 25 % in the primary HNC tumour uptake. Moreover, PVC enhanced the statistical significance of the results. CONCLUSION: FDG PET/CT uptake parameters are predictors of patients' outcome and can potentially identify patients with higher risk of treatment failure that could benefit from more aggressive approaches. Application of PVC is recommended for accurate measurement of PET parameters.

Machine learning on brain MRI data for differential diagnosis of Parkinson's disease and Progressive Supranuclear Palsy.

BACKGROUND: Supervised machine learning has been proposed as a revolutionary approach for identifying sensitive medical image biomarkers (or combination of them) allowing for automatic diagnosis of individual subjects. The aim of this work was to assess the feasibility of a supervised machine learning algorithm for the assisted diagnosis of patients with clinically diagnosed Parkinson's disease (PD) and Progressive Supranuclear Palsy (PSP). METHOD: Morphological T1-weighted Magnetic Resonance Images (MRIs) of PD patients (28), PSP patients (28) and healthy control subjects (28) were used by a supervised machine learning algorithm based on the combination of Principal Components Analysis as feature extraction technique and on Support Vector Machines as classification algorithm. The algorithm was able to obtain voxel-based morphological biomarkers of PD and PSP. RESULTS: The algorithm allowed individual diagnosis of PD versus controls, PSP versus controls and PSP versus PD with an Accuracy, Specificity and Sensitivity>90%. Voxels influencing classification between PD and PSP patients involved midbrain, pons, corpus callosum and thalamus, four critical regions known to be strongly involved in the pathophysiological mechanisms of PSP. COMPARISON WITH EXISTING METHODS: Classification accuracy of individual PSP patients was consistent with previous manual morphological metrics and with other supervised machine learning application to MRI data, whereas accuracy in the detection of individual PD patients was significantly higher with our classification method. CONCLUSIONS: The algorithm provides excellent discrimination of PD patients from PSP patients at an individual level, thus encouraging the application of computer-based diagnosis in clinical practice.

Pathway-based expression profile for breast cancer diagnoses.

Microarray experiments have made possible to identify breast cancer marker gene signatures. However, gene expression-based signatures present limitations because they do not consider metabolic role of the genes and are affected by genetic heterogeneity across patient cohorts. Considering the activity of entire pathways rather than the expression levels of individual genes can be a way to exceed these limits. We evaluated and compared five methods of pathway-level aggregation of gene expression data. Our results confirmed the important role of pathway expression profile in breast cancer diagnostic classification (accuracy >90%). However, although assessed on a limited number of samples and datasets, this study shows that using dissimilarity representation among patients does not improve the classification of pathway-based expression profiles.

Combination of gene expression and genome copy number alteration has a prognostic value for breast cancer.

Specific genome copy number alterations, such as deletions and amplifications are an important factor in tumor development and progression, and are also associated with changes in gene expression. By combining analyses of gene expression and genome copy number we identified genes as candidate biomarkers of BC which were validated as prognostic factors of the disease progression. These results suggest that the proposed combined approach may become a valuable method for BC prognosis.

A partial volume effect correction tailored for 18F-FDG-PET oncological studies.

We have developed, optimized, and validated a method for partial volume effect (PVE) correction of oncological lesions in positron emission tomography (PET) clinical studies, based on recovery coefficients (RC) and on PET measurements of lesion-to-background ratio (L/B m) and of lesion metabolic volume. An operator-independent technique, based on an optimised threshold of the maximum lesion uptake, allows to define an isocontour around the lesion on PET images in order to measure both lesion radioactivity uptake and lesion metabolic volume. RC are experimentally derived from PET measurements of hot spheres in hot background, miming oncological lesions. RC were obtained as a function of PET measured sphere-to-background ratio and PET measured sphere metabolic volume, both resulting from the threshold-isocontour technique. PVE correction of lesions of a diameter ranging from 10 mm to 40 mm and for measured L/B m from 2 to 30 was performed using measured RC curves tailored at answering the need to quantify a large variety of real oncological lesions by means of PET. Validation of the PVE correction method resulted to be accurate (>89%) in clinical realistic conditions for lesion diameter > 1 cm, recovering >76% of radioactivity for lesion diameter < 1 cm. Results from patient studies showed that the proposed PVE correction method is suitable and feasible and has an impact on a clinical environment.

The SRA protein UHRF1 promotes epigenetic crosstalks and is involved in prostate cancer progression.

Epigenetic silencing of tumour suppressor genes is an important mechanism involved in cell transformation and tumour progression. The Set and RING-finger-associated domain-containing protein UHRF1 might be an important link between different epigenetic pathways. Here, we report that UHRF1 is frequently overexpressed in human prostate tumours and has an important role in prostate cancer pathogenesis and progression. Analysis of human prostate cancer samples by microarrays and immunohistochemistry showed increased expression of UHRF1 in about half of the cases. Moreover, UHRF1 expression was associated with reduced overall survival after prostatectomy in patients with organ-confined prostate tumours (P < 0.0001). UHRF1 expression was negatively correlated with several tumour suppressor genes and positively with the histone methyltransferase (HMT) EZH2 both in prostate tumours and cell lines. UHRF1 knockdown reduced proliferation, clonogenic capability and anchorage-independent growth of prostate cancer cells. Depletion of UHRF1 resulted in reactivation of several tumour suppressor genes. Gene reactivation upon UHRF1 depletion was associated with changes in histone H3K9 methylation, acetylation and DNA methylation, and impaired binding of the H3K9 HMT Suv39H1 to the promoter of silenced genes. Co-immunoprecipitation experiments showed direct interaction between UHRF1 and Suv39H1. Our data support the notion that UHRF1, along with Suv39H1 and DNA methyltransferases, contributes to epigenetic gene silencing in prostate tumours. This could represent a parallel and convergent pathway to the H3K27 methylation catalyzed by EZH2 to synergistically promote inactivation of tumour suppressor genes. Deregulated expression of UHRF1 is involved in the prostate cancer pathogenesis and might represent a useful marker to distinguish indolent cancer from those at high risk of lethal progression.

A Decision Support System for the assisted diagnosis of brain tumors: a feasibility study for ¹8F-FDG PET preclinical studies.

Decision support systems for the assisted medical diagnosis offer the main feature of giving assessments which are poorly affected from arbitrary clinical reasoning. Aim of this work was to assess the feasibility of a decision support system for the assisted diagnosis of brain cancer, such approach presenting potential for early diagnosis of tumors and for the classification of the degree of the disease progression. For this purpose, a supervised learning algorithm combined with a pattern recognition method was developed and cross-validated in ¹8F-FDG PET studies of a model of a brain tumour implantation.

Two-dimensional vs three-dimensional imaging in whole body oncologic PET/CT: a Discovery-STE phantom and patient study.

AIM: To evaluate the performance of the positron emission tomography (PET)/computed tomography (CT) Discovery-STE (D-STE) scanner for lesion detectability in two-dimensional (2D) and three-dimensional (3D) acquisition. METHODS: A NEMA 2001 Image-Quality phantom with 11 lesions (7-37 mm in diameter) filled with a solution of 18F (lesion/background concentration ratio: 4.4) was studied. 2D and 3D PET scans were sequentially acquired (10 min each) in list mode (LM). Each scan was unlisted into 4, 3 and 2-min scans. Ten [18F]FDG PET oncological patient studies were also evaluated. Each patient underwent a 3D PET/CT whole body scan, followed by a 2D PET scan (4 min LM) and a 3D PET scan (4 min LM) over a single field of view. Both 2D and 3D scans were unlisted in 3 and 2-min scans. Data were evaluated quantitatively by calculating quality measurements and qualitatively by two physicians who judged lesion detectability compared to statistical variations in background activity. RESULTS: Quantitative and qualitative evaluations showed the superiority of 3D over 2D across all measures of quality. In particular, lesion detectability was better in 3D than in 2D at equal scan times and 3D acquisition provided images comparable in quality to 2D in approximately half the time. Interobserver variability was lower in evaluation of 3D scans and lesion shape and volume were better depicted. CONCLUSION: In oncological applications, the D-STE system demonstrated good performance in 2D and 3D acquisition, while 3D exhibited better image quality, data accuracy and consistency of lesion detectability, resulting in shorter scan times and higher patient throughput.

Using deconvolution to improve PET spatial resolution in OSEM iterative reconstruction.

OBJECTIVES: A novel approach to the PET image reconstruction is presented, based on the inclusion of image deconvolution during conventional OSEM reconstruction. Deconvolution is here used to provide a recovered PET image to be included as "a priori" information to guide OSEM toward an improved solution. METHODS: Deconvolution was implemented using the Lucy-Richardson (LR) algorithm: Two different deconvolution schemes were tested, modifying the conventional OSEM iterative formulation: 1) We built a regularizing penalty function on the recovered PET image obtained by deconvolution and included it in the OSEM iteration. 2) After each conventional global OSEM iteration, we deconvolved the resulting PET image and used this "recovered" version as the initialization image for the next OSEM iteration. Tests were performed on both simulated and acquired data. RESULTS: Compared to the conventional OSEM, both these strategies, applied to simulated and acquired data, showed an improvement in image spatial resolution with better behavior in the second case. In this way, small lesions, present on data, could be better discriminated in terms of contrast. CONCLUSIONS: Application of this approach to both simulated and acquired data suggests its efficacy in obtaining PET images of enhanced quality.

Early diagnosis of Alzheimer's disease using a grid implementation of statistical parametric mapping analysis.

A quantitative statistical analysis of perfusional medical images may provide powerful support to the early diagnosis for Alzheimer's Disease (AD). A Statistical Parametric Mapping algorithm (SPM), based on the comparison of the candidate with normal cases, has been validated by the neurological research community to quantify ipometabolic patterns in brain PET/SPECT studies. Since suitable "normal patient" PET/SPECT images are rare and usually sparse and scattered across hospitals and research institutions, the Data Grid distributed analysis paradigm ("move code rather than input data") is well suited for implementing a remote statistical analysis use case, described in the present paper. Different Grid environments (LCG, AliEn) and their services have been used to implement the above-described use case and tackle the challenging problems related to the SPM-based early AD diagnosis.

Automatic registration of PET and CT studies for clinical use in thoracic and abdominal conformal radiotherapy.

AIM: Implementation and validation of an automatic registration method based on mutual information (MI) for the integration of thoracic and abdominal positron emission tomography (PET)/computed tomography (CT) studies, with the purpose to facilitate in a clinical context the inclusion of PET metabolic information in conformal radiotherapy (RT). METHODS: Registration was obtained by modeling a rigid spatial transformation between CT and PET transmission studies. The registration method was based on Normalized Mutual Information (NMI), by iteratively transforming the PET volume, until its optimal alignment to the CT study is achieved, in correspondence of the maximum of NMI. To avoid entrapment in local maxima and to improve convergence speed we introduced a multiresolution scheme. Accuracy of the proposed approach was investigated in experimental data, relative to phantom and patient studies, acquired in conditions similar to clinical situations. RESULTS: In phantom studies the mean error in the 3D space is 3.6 mm (range 3-4 mm) in thoracic region and 3.2 mm (range 2.9-3.7 mm) in abdominal region, considerably less than PET spatial resolution. In patient studies the spatial mean error increases with respect to phantom studies (5.4 mm and 5.2 mm for thorax and abdomen, respectively) but remains comparable to the PET spatial resolution. The accuracy of spatial realignment was thus found adequate for the registration of PET/CT registration, if good patient repositioning was adopted. CONCLUSIONS: The proposed registration method, based on MI, was validated for the integration of PET/CT studies of patients candidate for thoracic and abdominal conformal RT. The method is automatic and provided with a user interface, thus suitable for clinical use.

Shared cortical anatomy for motor awareness and motor control.

In everyday life, the successful monitoring of behavior requires continuous updating of the effectiveness of motor acts; one crucial step is becoming aware of the movements one is performing. We studied the anatomical distribution of lesions in right-brain-damaged hemiplegic patients, who obstinately denied their motor impairment, claiming that they could move their paralyzed limbs. Denial was associated with lesions in areas related to the programming of motor acts, particularly Brodmann's premotor areas 6 and 44, motor area 4, and the somatosensory cortex. This association suggests that monitoring systems may be implemented within the same cortical network that is responsible for the primary function that has to be monitored.

A publicly accessible Monte Carlo database for validation purposes in emission tomography.

Monte Carlo (MC) methods provide ideal data sets to assess reconstruction and correction techniques in emission tomography (ET). Although several ET-dedicated MC codes are available, their use is hindered by the heavy computation burden required for high statistics simulations as well as by the need to adapt the code to the purpose of the individual user. In this work a publicly accessible database of MC-simulated ET data sets (the MC-ET database) was created and published on an Internet web site (http://www.ibfm.cnr.it/mcet/index.html), in order to provide MC-simulated data ready to be downloaded and used by researchers at different sites with similar evaluation purposes. At present, the MC-ET database provides direct access to MC-simulated raw data of unscattered, scattered and total events: (a) obtained by different MC codes, (b) relative to different radioactive sources, from simple geometrical phantoms to studies of normal and pathological subjects and (c) derived from different SPECT and PET scanners. The main features of the MC-ET data sets are: (a) validation by comparison with measured data, (b) classification according to pre-defined database characteristics, (c) common-use file format and (d) easy and free access and download.

Multi-modal medical image integration to optimize radiotherapy planning in lung cancer treatment.

This work presents a method for CT and PET image registration, and multi-modal analysis, to optimize radiotherapy planning in lung cancer treatment. The method relies on an image registration technique based on fiducial external markers to realign, spatially, PET images with the CT spatial reference system. The method was set up for clinical use in radiotherapy, allowing minimal modifications to be introduced in the management of patients undergoing radiation treatment. The accuracy of the registration technique was evaluated on patient studies in terms of Target Registration Error and was found to be less than 6.40 mm. The method was applied in the treatment planning of five patients affected by non-small-cell lung cancer, revealing the usefulness of PET/CT integration in delineating the extension of both the tumor mass and the tissues involved in the neoplastic process. Moreover, the functional information provided by PET often led to alterations in the treatment planning, changing the size and/or direction of radiation portals. The proposed method for PET/CT integration has been confirmed as being useful for optimizing radiotherapy planning in lung cancer treatment.