Performance Analysis of Six Semi-Automated Tumour Delineation Methods on [18F] Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography (FDG PET/CT) in Patients with Head and Neck Cancer.

Background. Head and neck cancer (HNC) is the seventh most common neoplastic disorder at the global level. Contouring HNC lesions on [18F] Fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT) scans plays a fundamental role for diagnosis, risk assessment, radiotherapy planning and post-treatment evaluation. However, manual contouring is a lengthy and tedious procedure which requires significant effort from the clinician. Methods. We evaluated the performance of six hand-crafted, training-free methods (four threshold-based, two algorithm-based) for the semi-automated delineation of HNC lesions on FDG PET/CT. This study was carried out on a single-centre population of n=103 subjects, and the standard of reference was manual segmentation generated by nuclear medicine specialists. Figures of merit were the Sørensen-Dice coefficient (DSC) and relative volume difference (RVD). Results. Median DSC ranged between 0.595 and 0.792, median RVD between -22.0% and 87.4%. Click and draw and Nestle's methods achieved the best segmentation accuracy (median DSC, respectively, 0.792 ± 0.178 and 0.762 ± 0.107; median RVD, respectively, -21.6% ± 1270.8% and -32.7% ± 40.0%) and outperformed the other methods by a significant margin. Nestle's method also resulted in a lower dispersion of the data, hence showing stronger inter-patient stability. The accuracy of the two best methods was in agreement with the most recent state-of-the art results. Conclusions. Semi-automated PET delineation methods show potential to assist clinicians in the segmentation of HNC lesions on FDG PET/CT images, although manual refinement may sometimes be needed to obtain clinically acceptable ROIs.

Form Factors as Potential Imaging Biomarkers to Differentiate Benign vs. Malignant Lung Lesions on CT Scans.

Indeterminate lung nodules detected on CT scans are common findings in clinical practice. Their correct assessment is critical, as early diagnosis of malignancy is crucial to maximise the treatment outcome. In this work, we evaluated the role of form factors as imaging biomarkers to differentiate benign vs. malignant lung lesions on CT scans. We tested a total of three conventional imaging features, six form factors, and two shape features for significant differences between benign and malignant lung lesions on CT scans. The study population consisted of 192 lung nodules from two independent datasets, containing 109 (38 benign, 71 malignant) and 83 (42 benign, 41 malignant) lung lesions, respectively. The standard of reference was either histological evaluation or stability on radiological followup. The statistical significance was determined via the Mann-Whitney U nonparametric test, and the ability of the form factors to discriminate a benign vs. a malignant lesion was assessed through multivariate prediction models based on Support Vector Machines. The univariate analysis returned four form factors (Angelidakis compactness and flatness, Kong flatness, and maximum projection sphericity) that were significantly different between the benign and malignant group in both datasets. In particular, we found that the benign lesions were on average flatter than the malignant ones; conversely, the malignant ones were on average more compact (isotropic) than the benign ones. The multivariate prediction models showed that adding form factors to conventional imaging features improved the prediction accuracy by up to 14.5 pp. We conclude that form factors evaluated on lung nodules on CT scans can improve the differential diagnosis between benign and malignant lesions.

Interval Fuzzy Model for Robust Aircraft IMU Sensors Fault Detection.

This paper proposes a data-based approach for a robust fault detection (FD) of the inertial measurement unit (IMU) sensors of an aircraft. Fuzzy interval models (FIMs) have been introduced for coping with the significant modeling uncertainties caused by poorly modeled aerodynamics. The proposed FIMs are used to compute robust prediction intervals for the measurements provided by the IMU sensors. Specifically, a nonlinear neural network (NN) model is used as central prediction of the sensor response while the uncertainty around the central estimation is captured by the FIM model. The uncertainty has been also modelled using a conventional linear Interval Model (IM) approach; this allows a quantitative evaluation of the benefits provided by the FIM approach. The identification of the IMs and of the FIMs was formalized as a linear matrix inequality (LMI) optimization problem using as cost function the (mean) amplitude of the prediction interval and as optimization variables the parameters defining the amplitudes of the intervals of the IMs and FIMs. Based on the identified models, FD validation tests have been successfully conducted using actual flight data of a P92 Tecnam aircraft by artificially injecting additive fault signals on the fault free IMU readings.

SPECT and PET serve as molecular imaging techniques and in vivo biomarkers for brain metastases.

Nuclear medicine techniques (single photon emission computerized tomography, SPECT, and positron emission tomography, PET) represent molecular imaging tools, able to provide in vivo biomarkers of different diseases. To investigate brain tumours and metastases many different radiopharmaceuticals imaged by SPECT and PET can be used. In this review the main and most promising radiopharmaceuticals available to detect brain metastases are reported. Furthermore the diagnostic contribution of the combination of SPECT and PET data with radiological findings (magnetic resonance imaging, MRI) is discussed.

Size measurement of lung nodules on CT: which diameter is most stable to inter-observer variability?

Indeterminate lung nodules detected on CT are common findings in the clinical practice, and the correct assessment of their size is critical for patient evaluation and management. We compared the stability of three definitions of nodule diameter (Feret's mean diameter, Martin's mean diameter and area-equivalent diameter) to inter-observer variability on a population of 336 solid nodules from 207 subjects. We found that inter-observer agreement was highest with Martin's mean diameter (intra-class correlation coefficient = 0.977, 95% Confidence interval = 0.977-0.978), followed by area-equivalent diameter (0.972, 0.971-0.973) and Feret's mean diameter (0.965, 0.964-0.966). The differences were statistically significant. In conclusion, although all the three diameter definitions achieved very good inter-observer agreement (ICC > 0.96), Martin's mean diameter was significantly better than the others. Future guidelines may consider adopting Martin's mean diameter as an alternative to the currently used Feret's (caliper) diameter for assessing the size of lung nodules on CT.

To what extent can artificial neural network support nuclear medicine?

Artificial Neural Networks (ANN) are computer programs that emulate the operation of a large number of processing units that mimic the fundamental mechanisms of the biological activity of nervous cells as well as their connections and interactions. As a human brain, an ANN has the ability to learn from the experience of general relations between variables and thus ANN are particularly suitable to capture the natural complexity of medical data. Today ANN are widely used as a tool for computer aided diagnosis. This editorial discusses to what extent ANN can support Nuclear Medicine.

Differential Diagnosis of Alzheimer Disease vs. Mild Cognitive Impairment Based on Left Temporal Lateral Lobe Hypomethabolism on 18F-FDG PET/CT and Automated Classifiers.

PURPOSE: We evaluate the ability of Artificial Intelligence with automatic classification methods applied to semi-quantitative data from brain 18F-FDG PET/CT to improve the differential diagnosis between Alzheimer Disease (AD) and Mild Cognitive Impairment (MCI). PROCEDURES: We retrospectively analyzed a total of 150 consecutive patients who underwent diagnostic evaluation for suspected AD (n = 67) or MCI (n = 83). All patients received brain 18F-FDG PET/CT according to the international guidelines, and images were analyzed both Qualitatively (QL) and Quantitatively (QN), the latter by a fully automated post-processing software that produced a z score metabolic map of 25 anatomically different cortical regions. A subset of n = 122 cases with a confirmed diagnosis of AD (n = 53) or MDI (n = 69) by 18-24-month clinical follow-up was finally included in the study. Univariate analysis and three automated classification models (classification tree -ClT-, ridge classifier -RC- and linear Support Vector Machine -lSVM-) were considered to estimate the ability of the z scores to discriminate between AD and MCI cases in. RESULTS: The univariate analysis returned 14 areas where the z scores were significantly different between AD and MCI groups, and the classification accuracy ranged between 74.59% and 76.23%, with ClT and RC providing the best results. The best classification strategy consisted of one single split with a cut-off value of ≈ -2.0 on the z score from temporal lateral left area: cases below this threshold were classified as AD and those above the threshold as MCI. CONCLUSIONS: Our findings confirm the usefulness of brain 18F-FDG PET/CT QL and QN analyses in differentiating AD from MCI. Moreover, the combined use of automated classifications models can improve the diagnostic process since its use allows identification of a specific hypometabolic area involved in AD cases in respect to MCI. This data improves the traditional 18F-FDG PET/CT image interpretation and the diagnostic assessment of cognitive disorders.

Comparative evaluation of conventional and deep learning methods for semi-automated segmentation of pulmonary nodules on CT.

BACKGROUND: Accurate segmentation of pulmonary nodules on computed tomography (CT) scans plays a crucial role in the evaluation and management of patients with suspicion of lung cancer (LC). When performed manually, not only the process requires highly skilled operators, but is also tiresome and time-consuming. To assist the physician in this task several automated and semi-automated methods have been proposed in the literature. In recent years, in particular, the appearance of deep learning has brought about major advances in the field. METHODS: Twenty-four (12 conventional and 12 based on deep learning) semi-automated-'one-click'-methods for segmenting pulmonary nodules on CT were evaluated in this study. The experiments were carried out on two datasets: a proprietary one (383 images from a cohort of 111 patients) and a public one (259 images from a cohort of 100). All the patients had a positive transcript for suspect pulmonary nodules. RESULTS: The methods based on deep learning clearly outperformed the conventional ones. The best performance [Sørensen-Dice coefficient (DSC)] in the two datasets was, respectively, 0.853 and 0.763 for the deep learning methods, and 0.761 and 0.704 for the traditional ones. CONCLUSIONS: Deep learning is a viable approach for semi-automated segmentation of pulmonary nodules on CT scans.

Impact of Lesion Delineation and Intensity Quantisation on the Stability of Texture Features from Lung Nodules on CT: A Reproducible Study.

Computer-assisted analysis of three-dimensional imaging data (radiomics) has received a lot of research attention as a possible means to improve the management of patients with lung cancer. Building robust predictive models for clinical decision making requires the imaging features to be stable enough to changes in the acquisition and extraction settings. Experimenting on 517 lung lesions from a cohort of 207 patients, we assessed the stability of 88 texture features from the following classes: first-order (13 features), Grey-level Co-Occurrence Matrix (24), Grey-level Difference Matrix (14), Grey-level Run-length Matrix (16), Grey-level Size Zone Matrix (16) and Neighbouring Grey-tone Difference Matrix (five). The analysis was based on a public dataset of lung nodules and open-access routines for feature extraction, which makes the study fully reproducible. Our results identified 30 features that had good or excellent stability relative to lesion delineation, 28 to intensity quantisation and 18 to both. We conclude that selecting the right set of imaging features is critical for building clinical predictive models, particularly when changes in lesion delineation and/or intensity quantisation are involved.

Comparison of two neural network classifiers in the differential diagnosis of essential tremor and Parkinson's disease by (123)I-FP-CIT brain SPECT.

PURPOSE: To contribute to the differentiation of Parkinson's disease (PD) and essential tremor (ET), we compared two different artificial neural network classifiers using (123)I-FP-CIT SPECT data, a probabilistic neural network (PNN) and a classification tree (ClT). METHODS: (123)I-FP-CIT brain SPECT with semiquantitative analysis was performed in 216 patients: 89 with ET, 64 with PD with a Hoehn and Yahr (H&Y) score of ≤2 (early PD), and 63 with PD with a H&Y score of ≥2.5 (advanced PD). For each of the 1,000 experiments carried out, 108 patients were randomly selected as the PNN training set, while the remaining 108 validated the trained PNN, and the percentage of the validation data correctly classified in the three groups of patients was computed. The expected performance of an "average performance PNN" was evaluated. In analogy, for ClT 1,000 classification trees with similar structures were generated. RESULTS: For PNN, the probability of correct classification in patients with early PD was 81.9±8.1% (mean±SD), in patients with advanced PD 78.9±8.1%, and in ET patients 96.6±2.6%. For ClT, the first decision rule gave a mean value for the putamen of 5.99, which resulted in a probability of correct classification of 93.5±3.4%. This means that patients with putamen values >5.99 were classified as having ET, while patients with putamen values <5.99 were classified as having PD. Furthermore, if the caudate nucleus value was higher than 6.97 patients were classified as having early PD (probability 69.8±5.3%), and if the value was <6.97 patients were classified as having advanced PD (probability 88.1%±8.8%). CONCLUSION: These results confirm that PNN achieved valid classification results. Furthermore, ClT provided reliable cut-off values able to differentiate ET and PD of different severities.

[123I]Metaiodobenzylguanidine (MIBG) Cardiac Scintigraphy and Automated Classification Techniques in Parkinsonian Disorders.

PURPOSE: To provide reliable and reproducible heart/mediastinum (H/M) ratio cut-off values for parkinsonian disorders using two machine learning techniques, Support Vector Machines (SVM) and Random Forest (RF) classifier, applied to [123I]MIBG cardiac scintigraphy. PROCEDURES: We studied 85 subjects, 50 with idiopathic Parkinson's disease, 26 with atypical Parkinsonian syndromes (P), and 9 with essential tremor (ET). All patients underwent planar early and delayed cardiac scintigraphy after [123I]MIBG (111 MBq) intravenous injection. Images were evaluated both qualitatively and quantitatively; the latter by the early and delayed H/M ratio obtained from regions of interest (ROIt1 and ROIt2) drawn on planar images. SVM and RF classifiers were finally used to obtain the correct cut-off value. RESULTS: SVM and RF produced excellent classification performances: SVM classifier achieved perfect classification and RF also attained very good accuracy. The better cut-off for H/M value was 1.55 since it remains the same for both ROIt1 and ROIt2. This value allowed to correctly classify PD from P and ET: patients with H/M ratio less than 1.55 were classified as PD while those with values higher than 1.55 were considered as affected by parkinsonism and/or ET. No difference was found when early or late H/M ratio were considered separately thus suggesting that a single early evaluation could be sufficient to obtain the final diagnosis. CONCLUSIONS: Our results evidenced that the use of SVM and CT permitted to define the better cut-off value for H/M ratios both in early and in delayed phase thus underlining the role of [123I]MIBG cardiac scintigraphy and the effectiveness of H/M ratio in differentiating PD from other parkinsonism or ET. Moreover, early scans alone could be used for a reliable diagnosis since no difference was found between early and late. Definitely, a larger series of cases is needed to confirm this data.

Value of Shape and Texture Features from 18F-FDG PET/CT to Discriminate between Benign and Malignant Solitary Pulmonary Nodules: An Experimental Evaluation.

In this paper, we investigate the role of shape and texture features from 18F-FDG PET/CT to discriminate between benign and malignant solitary pulmonary nodules. To this end, we retrospectively evaluated cross-sectional data from 111 patients (64 males, 47 females, age = 67.5 ± 11.0) all with histologically confirmed benign (n=39) or malignant (n=72) solitary pulmonary nodules. Eighteen three-dimensional imaging features, including conventional, texture, and shape features from PET and CT were tested for significant differences (Wilcoxon-Mann-Withney) between the benign and malignant groups. Prediction models based on different feature sets and three classification strategies (Classification Tree, k-Nearest Neighbours, and Naïve Bayes) were also evaluated to assess the potential benefit of shape and texture features compared with conventional imaging features alone. Eight features from CT and 15 from PET were significantly different between the benign and malignant groups. Adding shape and texture features increased the performance of both the CT-based and PET-based prediction models with overall accuracy gain being 3.4-11.2 pp and 2.2-10.2 pp, respectively. In conclusion, we found that shape and texture features from 18F-FDG PET/CT can lead to a better discrimination between benign and malignant lung nodules by increasing the accuracy of the prediction models by an appreciable margin.

Classification Model to Estimate MIB-1 (Ki 67) Proliferation Index in NSCLC Patients Evaluated With 18F-FDG-PET/CT.

BACKGROUND/AIM: Proliferation biomarkers such as MIB-1 are strong predictors of clinical outcome and response to therapy in patients with non-small-cell lung cancer, but they require histological examination. In this work, we present a classification model to predict MIB-1 expression based on clinical parameters from positron emission tomography. PATIENTS AND METHODS: We retrospectively evaluated 78 patients with histology-proven non-small-cell lung cancer (NSCLC) who underwent 18F-FDG-PET/CT for clinical examination. We stratified the population into a low and high proliferation group using MIB-1=25% as cut-off value. We built a predictive model based on binary classification trees to estimate the group label from the maximum standardized uptake value (SUVmax) and lesion diameter. RESULTS: The proposed model showed ability to predict the correct proliferation group with overall accuracy >82% (78% and 86% for the low- and high-proliferation group, respectively). CONCLUSION: Our results indicate that radiotracer activity evaluated via SUVmax and lesion diameter are correlated with tumour proliferation index MIB-1.

Texture Analysis on [18F]FDG PET/CT in Non-Small-Cell Lung Cancer: Correlations Between PET Features, CT Features, and Histological Types.

PURPOSE: The study aims to investigate the correlations between positron emission tomography (PET) texture features, X-ray computed tomography (CT) texture features, and histological subtypes in non-small-cell lung cancer evaluated with 2-deoxy-2-[18F]fluoro-D-glucose PET/CT. PROCEDURES: We retrospectively evaluated the baseline PET/CT scans of 81 patients with histologically proven non-small-cell lung cancer. Feature extraction and statistical analysis were carried out on the Matlab platform (MathWorks, Natick, USA). RESULTS: Intra-CT correlation analysis revealed a strong positive correlation between volume of the lesion (CTvol) and maximum density (CTmax), and between kurtosis (CTkrt) and maximum density (CTmax). A moderate positive correlation was found between volume (CTvol) and average density (CTmean), and between kurtosis (CTkrt) and average density (CTmean). Intra-PET analysis identified a strong positive correlation between the radiotracer uptake (SUVmax, SUVmean) and its degree of variability/disorder throughout the lesion (SUVstd, SUVent). Conversely, there was a strong negative correlation between the uptake (SUVmax, SUVmean) and its degree of uniformity (SUVuni). There was a positive moderate correlation between the metabolic tumor volume (MTV) and radiotracer uptake (SUVmax, SUVmean). Inter (PET-CT) correlation analysis identified a very strong positive correlation between the volume of the lesion at CT (CTvol) and the metabolic volume (MTV), a moderate positive correlation between average tissue density (CTmean) and radiotracer uptake (SUVmax, SUVmean), and between kurtosis at CT (CTkrt) and metabolic tumor volume (MTV). Squamous cell carcinomas had larger volume higher uptake, stronger PET variability and lower uniformity than the other subtypes. By contrast, adenocarcinomas exhibited significantly lower uptake, lower variability and higher uniformity than the other subtypes. CONCLUSIONS: Significant associations emerged between PET features, CT features, and histological type in NSCLC. Texture analysis on PET/CT shows potential to differentiate between histological types in patients with non-small-cell lung cancer.

Evaluation of Shape and Textural Features from CT as Prognostic Biomarkers in Non-small Cell Lung Cancer.

BACKGROUND/AIM: We retrospectively investigated the prognostic potential (correlation with overall survival) of 9 shape and 21 textural features from non-contrast-enhanced computed tomography (CT) in patients with non-small-cell lung cancer. MATERIALS AND METHODS: We considered a public dataset of 203 individuals with inoperable, histologically- or cytologically-confirmed NSCLC. Three-dimensional shape and textural features from CT were computed using proprietary code and their prognostic potential evaluated through four different statistical protocols. RESULTS: Volume and grey-level run length matrix (GLRLM) run length non-uniformity were the only two features to pass all four protocols. Both features correlated negatively with overall survival. The results also showed a strong dependence on the evaluation protocol used. CONCLUSION: Tumour volume and GLRLM run-length non-uniformity from CT were the best predictor of survival in patients with non-small-cell lung cancer. We did not find enough evidence to claim a relationship with survival for the other features.

Role of Artificial Intelligence Techniques (Automatic Classifiers) in Molecular Imaging Modalities in Neurodegenerative Diseases.

Artificial Intelligence (AI) is a very active Computer Science research field aiming to develop systems that mimic human intelligence and is helpful in many human activities, including Medicine. In this review we presented some examples of the exploiting of AI techniques, in particular automatic classifiers such as Artificial Neural Network (ANN), Support Vector Machine (SVM), Classification Tree (ClT) and ensemble methods like Random Forest (RF), able to analyze findings obtained by positron emission tomography (PET) or single-photon emission tomography (SPECT) scans of patients with Neurodegenerative Diseases, in particular Alzheimer's Disease. We also focused our attention on techniques applied in order to preprocess data and reduce their dimensionality via feature selection or projection in a more representative domain (Principal Component Analysis - PCA - or Partial Least Squares - PLS - are examples of such methods); this is a crucial step while dealing with medical data, since it is necessary to compress patient information and retain only the most useful in order to discriminate subjects into normal and pathological classes. Main literature papers on the application of these techniques to classify patients with neurodegenerative disease extracting data from molecular imaging modalities are reported, showing that the increasing development of computer aided diagnosis systems is very promising to contribute to the diagnostic process.

Diagnostic accuracy of Parkinson disease by support vector machine (SVM) analysis of 123I-FP-CIT brain SPECT data: implications of putaminal findings and age.

Brain single-photon-emission-computerized tomography (SPECT) with I-ioflupane (I-FP-CIT) is useful to diagnose Parkinson disease (PD). To investigate the diagnostic performance of I-FP-CIT brain SPECT with semiquantitative analysis by Basal Ganglia V2 software (BasGan), we evaluated semiquantitative data of patients with suspect of PD by a support vector machine classifier (SVM), a powerful supervised classification algorithm.I-FP-CIT SPECT with BasGan analysis was performed in 90 patients with suspect of PD showing mild symptoms (bradykinesia-rigidity and mild tremor). PD was confirmed in 56 patients, 34 resulted non-PD (essential tremor and drug-induced Parkinsonism). A clinical follow-up of at least 6 months confirmed diagnosis. To investigate BasGan diagnostic performance we trained SVM classification models featuring different descriptors using both a "leave-one-out" and a "five-fold" method. In the first study we used as class descriptors the semiquantitative radiopharmaceutical uptake values in the left (L) and right (R) putamen (P) and in the L and R caudate nucleus (C) for a total of 4 descriptors (CL, CR, PL, PR). In the second study each patient was described only by CL and CR, while in the third by PL and PR descriptors. Age was added as a further descriptor to evaluate its influence in the classification performance.I-FP-CIT SPECT with BasGan analysis reached a classification performance higher than 73.9% in all the models. Considering the "Leave-one-out" method, PL and PR were better predictors (accuracy of 91% for all patients) than CL and CR descriptors; using PL, PR, CL, and CR diagnostic accuracy was similar to that of PL and PR descriptors in the different groups. Adding age as a further descriptor accuracy improved in all the models. The best results were obtained by using all the 5 descriptors both in PD and non-PD subjects (CR and CL + PR and PL + age = 96.4% and 94.1%, respectively). Similar results were observed for the "five-fold" method. I-FP-CIT SPECT with BasGan analysis using SVM classifier was able to diagnose PD. Putamen was the most discriminative descriptor for PD and the patient age influenced the classification accuracy.