Search

Background: This study investigated alternative, non-invasive methods for human papillomavirus (HPV) detection in head and neck cancers (HNCs). We compared two approaches: analyzing computed tomography (CT) scans with a Deep Learning (DL) model and using radiomic features extracted from CT images with machine learning (ML) models., Methods: Fifty patients with histologically confirmed HNC were included. We first trained a modified ResNet-18 DL model on CT data to predict HPV status. Next, radiomic features were extracted from manually segmented regions of interest near the oropharynx and used to train four ML models (K-Nearest Neighbors, logistic regression, decision tree, random forest) for the same purpose., Results: The CT-based model achieved the highest accuracy (90%) in classifying HPV status. Among the ML models, K-Nearest Neighbors performed best (80% accuracy). Weighted Ensemble methods combining the CT-based model with each ML model resulted in moderate accuracy improvements (70-90%)., Conclusions: Our findings suggest that CT scans analyzed by DL models hold promise for non-invasive HPV detection in HNC. Radiomic features, while less accurate in this study, offer a complementary approach. Future research should explore larger datasets and investigate the potential of combining DL and radiomic techniques.

The study investigates the efficiency of integrating Machine Learning (ML) in clinical practice for diagnosing solitary pulmonary nodules' (SPN) malignancy. Patient data had been recorded in the Department of Nuclear Medicine, University Hospital of Patras, in Greece. A dataset comprising 456 SPN characteristics extracted from CT scans, the SUVmax score from the PET examination, and the ultimate outcome (benign/malignant), determined by patient follow-up or biopsy, was used to build the ML classifier. Two medical experts provided their malignancy likelihood scores, taking into account the patient's clinical condition and without prior knowledge of the true label of the SPN. Incorporating human assessments into ML model training improved diagnostic efficiency by approximately 3%, highlighting the synergistic role of human judgment alongside ML. Under the latter setup, the ML model had an accuracy score of 95.39% (CI 95%: 95.29-95.49%). While ML exhibited swings in probability scores, human readers excelled in discerning ambiguous cases. ML outperformed the best human reader in challenging instances, particularly in SPNs with ambiguous probability grades, showcasing its utility in diagnostic grey zones. The best human reader reached an accuracy of 80% in the grey zone, whilst ML exhibited 89%. The findings underline the collaborative potential of ML and human expertise in enhancing SPN characterization accuracy and confidence, especially in cases where diagnostic certainty is elusive. This study contributes to understanding how integrating ML and human judgement can optimize SPN diagnostic outcomes, ultimately advancing clinical decision-making in PET/CT screenings.

Fuzzy Cognitive Maps (FCMs) have become an invaluable tool for healthcare providers because they can capture intricate associations among variables and generate precise predictions. FCMs have demonstrated their utility in diverse medical applications, from disease diagnosis to treatment planning and prognosis prediction. Their ability to model complex relationships between symptoms, biomarkers, risk factors, and treatments has enabled healthcare providers to make informed decisions, leading to better patient outcomes. This review article provides a thorough synopsis of using FCMs within the medical domain. A systematic examination of pertinent literature spanning the last two decades forms the basis of this overview, specifically delineating the diverse applications of FCMs in medical realms, including decision-making, diagnosis, prognosis, treatment optimisation, risk assessment, and pharmacovigilance. The limitations inherent in FCMs are also scrutinised, and avenues for potential future research and application are explored.

Deep learning (DL) is a well-established pipeline for feature extraction in medical and nonmedical imaging tasks, such as object detection, segmentation, and classification. However, DL faces the issue of explainability, which prohibits reliable utilisation in everyday clinical practice. This study evaluates DL methods for their efficiency in revealing and suggesting potential image biomarkers. Eleven biomedical image datasets of various modalities are utilised, including SPECT, CT, photographs, microscopy, and X-ray. Seven state-of-the-art CNNs are employed and tuned to perform image classification in tasks. The main conclusion of the research is that DL reveals potential biomarkers in several cases, especially when the models are trained from scratch in domains where low-level features such as shapes and edges are not enough to make decisions. Furthermore, in some cases, device acquisition variations slightly affect the performance of DL models.

Purpose: This paper reviews recent applications of Generative Adversarial Networks (GANs) in Positron Emission Tomography (PET) imaging. Recent advances in Deep Learning (DL) and GANs catalysed the research of their applications in medical imaging modalities. As a result, several unique GAN topologies have emerged and been assessed in an experimental environment over the last two years., Methods: The present work extensively describes GAN architectures and their applications in PET imaging. The identification of relevant publications was performed via approved publication indexing websites and repositories. Web of Science, Scopus, and Google Scholar were the major sources of information., Results: The research identified a hundred articles that address PET imaging applications such as attenuation correction, de-noising, scatter correction, removal of artefacts, image fusion, high-dose image estimation, super-resolution, segmentation, and cross-modality synthesis. These applications are presented and accompanied by the corresponding research works., Conclusion: GANs are rapidly employed in PET imaging tasks. However, specific limitations must be eliminated to reach their full potential and gain the medical community's trust in everyday clinical practice., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

Background: Parathyroid proliferative disorder encompasses a wide spectrum of diseases, including parathyroid adenoma (PTA), parathyroid hyperplasia, and parathyroid carcinoma. Imaging modalities that deliver their results preoperatively help in the localisation of parathyroid glands (PGs) and assist in surgery. Artificial intelligence and, more specifically, image detection methods, can assist medical experts and reduce the workload in their everyday routine., Methods: The present study employs an innovative CNN topology called ParaNet, to analyse early MIBI, late MIBI, and TcO4 thyroid scan images simultaneously to perform first-level discrimination between patients with abnormal PGs (aPG) and patients with normal PGs (nPG). The study includes 632 parathyroid scans., Results: ParaNet exhibits a top performance, reaching an accuracy of 96.56% in distinguishing between aPG and nPG scans. Its sensitivity and specificity are 96.38% and 97.02%, respectively. PPV and NPV values are 98.76% and 91.57%, respectively., Conclusions: The proposed network is the first to introduce the automatic discrimination of PG and nPG scans acquired by scintigraphy with 99m Tc-sestamibi (MIBI). This methodology could be applied to the everyday routine of medics for real-time evaluation or educational purposes.

Early and automatic diagnosis of Solitary Pulmonary Nodules (SPN) in Computed Tomography (CT) chest scans can provide early treatment for patients with lung cancer, as well as doctor liberation from time-consuming procedures. The purpose of this study is the automatic and reliable characterization of SPNs in CT scans extracted from Positron Emission Tomography and Computer Tomography (PET/CT) system. To achieve the aforementioned task, Deep Learning with Convolutional Neural Networks (CNN) is applied. The strategy of training specific CNN architectures from scratch and the strategy of transfer learning, by utilizing state-of-the-art pre-trained CNNs, are compared and evaluated. To enhance the training sets, data augmentation is performed. The publicly available database of CT scans, named as Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI), is also utilized to further expand the training set and is added to the PET/CT dataset. The results highlight the effectiveness of transfer learning and data augmentation for the classification task of small datasets. The best accuracy obtained on the PET/CT dataset reached 94%, utilizing a modification proposal of a state-of-the-art CNN, called VGG16, and enhancing the training set with LIDC-IDRI dataset. Besides, the proposed modification outperforms in terms of sensitivity several similar researches, which exploit the benefits of transfer learning. Overview of the experiment setup. The two datasets containing nodule representations are combined to evaluate the effectiveness of transfer learning over the traditional approach of training Convolutional Neural Networks from scratch.

Purpose: Accurate detection and treatment of Coronary Artery Disease is mainly based on invasive Coronary Angiography, which could be avoided provided that a robust, non-invasive detection methodology emerged. Despite the progress of computational systems, this remains a challenging issue. The present research investigates Machine Learning and Deep Learning methods in competing with the medical experts' diagnostic yield. Although the highly accurate detection of Coronary Artery Disease, even from the experts, is presently implausible, developing Artificial Intelligence models to compete with the human eye and expertise is the first step towards a state-of-the-art Computer-Aided Diagnostic system., Methods: A set of 566 patient samples is analysed. The dataset contains Polar Maps derived from scintigraphic Myocardial Perfusion Imaging studies, clinical data, and Coronary Angiography results. The latter is considered as reference standard. For the classification of the medical images, the InceptionV3 Convolutional Neural Network is employed, while, for the categorical and continuous features, Neural Networks and Random Forest classifier are proposed., Results: The research suggests that an optimal strategy competing with the medical expert's accuracy involves a hybrid multi-input network composed of InceptionV3 and a Random Forest. This method matches the expert's accuracy, which is 79.15% in the particular dataset., Conclusion: Image classification using deep learning methods can cooperate with clinical data classification methods to enhance the robustness of the predicting model, aiming to compete with the medical expert's ability to identify Coronary Artery Disease subjects, from a large scale patient dataset., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2021 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.)

Objective: To investigate a deep learning technique, more specifically state-of-the-art convolutional neural networks (CNN), for automatic characterization of polar maps derived from myocardial perfusion imaging (MPI) studies for the diagnosis of coronary artery disease., Subjects and Methods: Stress and rest polar maps corresponding to 216 patient cases from the database of the department of Nuclear Medicine of our institution were analyzed. Both attenuation-corrected (AC) and non-corrected (NAC) images were included. All patients were subjected to invasive coronary angiography within 60 days from MPI. As the initial dataset of this study was small to train a deep learning model from scratch, two strategies were followed. The first is called transfer learning. For this, we employed the state-of-the-art CNN called VGG16, which has been broadly exploited in medical imaging classification tasks. The second strategy involves data augmentation, which is achieved by the rotation of the polar maps, to expand the training set. We evaluated VGG16 with 10-fold cross-validation on the original set of images performing separate experiments for AC and NAC polar maps, as well as for their combination. The results were compared to the standard semi-quantitative polar map analysis based on summed stress and summed difference scores, as well as to the medical experts' diagnostic yield., Results: With reference to the findings of coronary angiography, VGG16 achieved an accuracy of 74.53%, sensitivity 75.00% and specificity 73.43% when the AC and NAC polar maps were incorporated into one single image set. Respective figures of MPI interpretation by experienced Nuclear Medicine physicians were 75.00%, 76.97% and 70.31%. The accuracy of semi-quantitative polar map analysis was lower, 66.20% and 64.81% for AC and NAC technique, respectively., Conclusion: The proposed deep learning model with data augmentation techniques performed better than the conventional semi-quantitative polar map analysis and competed with doctor's expertise in this particular patient cohort and image set. The model could potentially serve as an assisting tool to support interpretation of MPI studies or could be used for teaching purposes.

Purpose: A systematic meta-analysis of published studies on the diagnostic accuracy of presynaptic dopaminergic imaging with ¹²³I-FP-CIT (DaTSCAN) in dementia with Lewy bodies (DLB)., Methods: We included (a) studies in which DaTSCAN was performed in cases of diagnostic uncertainty to differentiate between DLB and non-DLB dementia and (b) studies of patients with already established diagnoses of DLB, non-DLB dementia, or normalcy, against which the diagnostic accuracy of DaTSCAN was tested. We applied fixed-effects Mantel-Haenszel and hierarchical logistic regression models for meta-analysis of the diagnostic test's accuracy. The pooled sensitivity, specificity and diagnostic odds ratio (DOR) were estimated. Heterogeneity among studies was estimated by calculating the corresponding I² statistic. We tested for publication or other bias by creating the respective funnel plot., Results: Four studies, with a total of 419 subjects, were judged to be suitable for meta-analysis. According to hierarchical models, the estimated pooled sensitivity of DaTSCAN in differentiating DLB versus no DLB was 86.5% [95% Confidence Intervals (CI): 72-94.1%], the specificity was 93.6% (95% CI: 88.5-96.6%), and the DOR was 94.2 (95% CI: 25.7-345). The Mantel-Haenszel estimate of overall DOR was calculated at 48.95 (95% CI: 26.16-91.59). There was some heterogeneity among studies (I² = 45%), but this was not found statistically significant (χ² test for heterogeneity p value = 0.142). Funnel plot analysis showed no significant bias., Conclusion: Allowing for the small number of studies in the meta-analysis, our results showed high diagnostic accuracy of DaTSCAN in DLB diagnosis, especially in terms of specificity., (Copyright © 2011 Elsevier Ltd. All rights reserved.)

This paper is a critical review of the literature on NET radionuclide therapy with (111)In-DTPA(0)-octreotide (Octreoscan) and (131)I-MIBG, focusing on efficacy and toxicity. Some potential future applications and new candidate therapeutic agents are also mentioned. Octreoscan has been a pioneering agent for somatostatin receptor radionuclide therapy. It has achieved symptomatic responses and disease stabilization, but it is now outperformed by the corresponding β-emitter agents (177)Lu-DOTATATE and (90)Y-DOTATOC. (131)I-MIBG is the radionuclide therapy of choice for inoperable or metastatic phaeochromocytomas/paragangliomas, which avidly concentrate this tracer via the noradrenaline transporter. Symptomatic, biochemical and tumour morphological response rates of 50-89%, 45-74% and 27-47%, respectively, have been reported. (131)I-MIBG is a second-line radiopharmaceutical for treatment of enterochromaffin carcinoids, mainly offering the benefit of amelioration of hormone-induced symptoms. High specific activity, non-carrier-added (131)I-MIBG and meta-astato((211)At)-benzylguanidine (MABG) are tracers with potential for enhanced therapeutic efficacy, yet their integration into clinical practice awaits further exploration. Amongst other promising agents, radiolabelled exendin analogues show potential for imaging and possibly therapy of insulinomas, while preclinical studies are currently evaluating DOTA peptides targeting the CCK-2/gastrin receptors that are overexpressed by medullary thyroid carcinoma cells.

The purpose of this study was to examine if factors of the external operating environment can explain differences in technical efficiency derived from Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA). In a sample of 124 dialysis facilities, technical efficiency was compared according to ownership, region, years in operation and size. With second-stage Tobit regression, DEA and SFA efficiency was regressed against these environmental factors to determine their potential for predicting technical efficiency, as well as the efficiency differences between the two frontier methods. DEA expectedly generated lower mean efficiency scores than SFA (68.2% vs. 79.4%, P < 0.001), due to the "random effects" term computed by the latter, in addition to "true" inefficiency. This finding was consistent for the subgroups formed on the basis of the environmental factors. Half the variation in the DEA-SFA efficiency differences was explained by environmental factors. This suggests that in addition to market instabilities, luck, and other related phenomena, decision-makers in their effort to determine optimal resource allocation, should point their attention to the potentially useful insight provided by environmental factors.

Unlabelled: The aim of our study was to evaluate prospectively the diagnostic performance and prognostic significance of (18)F-FDG PET/CT in comparison with (123)I-metaiodobenzylguanidine ((123)I-MIBG) imaging in patients with high-risk neuroblastoma., Methods: Twenty-eight patients with refractory or relapsed high-risk neuroblastoma (16 male and 12 female patients; age range, 2-45 y; median age, 7.5 y) were simultaneously evaluated with (18)F-FDG PET/CT and (123)I-MIBG imaging before treatment with high-dose (131)I-MIBG. We compared the 2 methods in mapping tumor load, according to the extent of disease and intensity of positive lesions identified in each patient. Separate comparisons were performed for the soft-tissue and bone-bone marrow components of tumor burden. Survival analysis was performed to assess the prognostic significance of (18)F-FDG and (123)I-MIBG imaging parameters., Results: (18)F-FDG PET/CT results were positive in 24 of 28 (86%) patients, whereas (123)I-MIBG imaging results were positive in all patients. (18)F-FDG was superior in mapping tumor load in 4 of 28 (14%) patients, whereas (123)I-MIBG was better in 12 of 28 (43%) patients. In the remaining 12 (43%) patients, no major differences were noted between the 2 modalities. (18)F-FDG PET/CT missed 5 cases of bone-bone marrow disease, 4 cases of soft-tissue disease, and 6 cases of skull involvement that were positive on (123)I-MIBG scans. Cox regression and Kaplan-Meier survival curves showed that the group of patients (4/28) in whom (18)F-FDG was superior to (123)I-MIBG had a significantly lower survival rate than the others. Tumoral avidity for (18)F-FDG (maximum standardized uptake value) and extent of (18)F-FDG-avid bone-bone marrow disease were identified as adverse prognostic factors., Conclusion: (123)I-MIBG imaging is superior to (18)F-FDG PET/CT in the assessment of disease extent in high-risk neuroblastoma. However, (18)F-FDG PET/CT has significant prognostic implications in these patients.

The aim of these guidelines is to assist nuclear medicine physicians in recommending, performing, reporting and interpreting the results of somatostatin (SST) receptor PET/CT imaging using 68Ga-DOTA-conjugated peptides, analogues of octreotide, that bind to SST receptors. This imaging modality should not be regarded as the only approach to visualizing tumours expressing SST receptors or as excluding other imaging modalities useful for obtaining comparable results. The corresponding guidelines of 111In-pentetreotide scintigraphy imaging have been considered and partially integrated with this text. The same has been done with the relevant and recent literature in this field and the final result has been discussed by distinguished experts.

In any production unit, the ability to achieve technical efficiency is influenced by characteristics of the external operating environment. This study uses the Greek dialysis sector to employ a previously reported frontier procedure to obtain a measure of managerial inefficiency that controls for exogenous features. The sample consisted of 124 dialysis facilities. Two inputs -nursing staff and dialysis machines- and one output -dialysis sessions--were used in an input-oriented, variable-returns-to-scale DEA model. Input slacks were regressed against environmental characteristics such as ownership, location, operating years and facility size, and parameter estimates were used to adjust primary input data. New efficiency scores were generated to measure managerial inefficiency. Older, public, regional facilities were operating under unfavorable circumstances, whereas newer, private, Athens-based facilities under favorable conditions. This respectively generated lower and higher efficiency scores than would have been attained on a level "playing field". After adjustment, scores reflected only management inefficiency and could be compared fairly. This study emphasizes the importance of efficiency comparisons, which take into account external conditions beyond the influence ofmanagement, as these have been shown to under- or overstate true management inefficiency.

Purpose: This was a retrospective study to detect and map the extent of disease in recurrent medullary thyroid carcinoma (MTC) using the novel PET somatostatin analogue (68)Ga-DOTATATE and conventional (18)F-FDG positron emission tomography/computed tomography (PET/CT)., Methods: Eighteen patients (13 men, 5 women, median age: 54 years) who had previously been operated on for MTC and presented with biochemical (raised calcitonin levels) and/or imaging evidence of recurrence underwent both (68)Ga-DOTATATE and (18)F-FDG PET/CT within a maximum interval of 4 weeks (median interval of 1 week). (68)Ga-DOTATATE- and (18)F-FDG-avid lesions were recorded per patient as well as per region in six distinct regions: (1) thyroid bed-local recurrence, (2) cervical lymph nodes, (3) mediastinum, (4) lungs, (5) liver and (6) bones. The (68)Ga-DOTATATE and (18)F-FDG PET/CT findings were classified as positive or negative on visual interpretation. These findings were further characterised as concordant or discordant, depending on whether there was agreement or discrepancy in imaging with the two radiotracers. A separate analysis of the unenhanced CT component of the examination was performed. Verification of the lesions was achieved by histopathological analysis, further imaging studies and clinical follow-up., Results: (68)Ga-DOTATATE PET/CT imaging achieved disease detection in 13 of 18 and (18)F-FDG PET/CT in 14 of 18 patients. These results corresponded to per-patient sensitivities of 72.2% [95% confidence interval (CI): 46.4-89.3%] for (68)Ga-DOTATATE versus 77.8% (95% CI: 51.9-92.6%) for (18)F-FDG (non-significant difference). (18)F-FDG revealed a total of 28 metastatic MTC regions and (68)Ga-DOTATATE 23 regions. In ten patients a discordant tracer pattern of per-region and/or per-lesion distribution of recurrent disease was observed, while in four patients a concordant pattern was noted (no lesions were detected by either modality in the remaining four patients)., Conclusion: Neither (18)F-FDG nor (68)Ga-DOTATATE PET/CT can fully map the extent of disease in patients with recurrent MTC, although (18)F-FDG PET/CT may identify more lesions. However, (68)Ga-DOTATATE PET/CT can be a useful complementary imaging tool and may identify patients suitable for consideration of targeted radionuclide somatostatin analogue therapy.

Objective: Left ventricular function is prognostically important. Our aim was to validate different algorithms' measurements with rubidium-82 PET, using computed tomography (CT) acquired simultaneously on hybrid imaging., Methods: Fifty patients (33 men, 17 women, mean age 59 years SD 12) referred for coronary artery disease evaluation underwent rubidium-82 PET myocardial perfusion scintigraphy and 64-slice CT coronary angiography simultaneously on hybrid PET/CT. Left ventricular ejection fraction, end-systolic volume and end-diastolic volume from quantitative gated PET at rest, were calculated using quantitative gated single-photon emission computed tomography (QGS), Emory Cardiac Toolbox (ECT), Myometrix and CardIQ Physio software and then compared with CT. The paired t-test and Bland-Altman plots were used for analysis., Results: There was no significant difference between CT and both QGS and ECT for calculating ESV. Otherwise there were significant differences (t-test, P<0.05) between CT coronary angiography and all other software algorithms in calculating systolic and diastolic volumes. Bland-Altman analysis revealed a bias towards underestimating end-diastolic volume with mean differences of 32.3, 34.8, 42.6 and 35.7 ml for QGS, ECT, Myometrix and CardIQ Physio, respectively. The limits of agreement (mean bias+/-1.96 SD) for left ventricular ejection fraction (%) were 8.5+/-21.0, 12.9+/-21.0, 3.5+/-18.9 and 5.6+/-17.1, for QGS, ECT, Myometrix and CardIQ Physio, respectively., Conclusion: The systematic error (or mean bias) of the Myometrix algorithm is the smallest, and CardIQ Physio shows the least statistical error (or limits of agreement) when calculating ejection fraction from rubidium-82 myocardial perfusion PET, compared with QGS and ECT. The wide limits of agreement between the different algorithms mean that they are not interchangeable.

Background and Aim: (99m)Tc-depreotide is a (99m)Tc-labelled somatostatin analogue, with high affinity for the 2, 3 and 5 subtypes of somatostatin receptors. These particular receptors are over-expressed on the surface of activated leucocytes, which mediate inflammatory response. Based on this property this study tried to investigate whether (99m)Tc-depreotide scintigraphy could be a useful complementary method in the investigation of bone infection and inflammation., Methods: Twenty-three patients, who were investigated for probable osteomyelitis, underwent three-phase bone scintigraphy followed by (99m)Tc-depreotide scintigraphy. Clinical and laboratory findings, complementary imaging procedures, clinical follow-up and bone biopsy established the final diagnosis. (99m)Tc-depreotide scintigraphy was performed 3 h after the intravenous administration of 555-740 MBq of the radiopharmaceutical. Scintigraphic images were, at first, blindly interpreted alone and then in comparative assessment with bone scans., Results: (99m)Tc-depreotide was positive in 12/12 cases of active osteomyelitis, one case of recent femoral head osteonecrosis and 6/9 rheumatoid arthritis sites. Negative (99m)Tc-depreotide scans were acquired in five cases of 'no-inflammation' (an uncomplicated fracture, an aseptic loosening of prosthesis, an old osteonecrosis, a healed and a successfully treated osteomyelitis), as well as in 14/14 total sites of degenerative arthritis-osteoarthropathy. In five cases (septic arthritis, periodontal and soft tissue infections) (99m)Tc-depreotide was positive, though spatially discordant with bone scintigraphy, delineating precisely the focus of infection., Conclusion: (99m)Tc-depreotide can be a useful complementary imaging method in the evaluation of bone infection and inflammation. Its combination with three-phase bone scintigraphy seems to be accurate in localizing the infection foci and determining the activity of the inflammatory processes.