Search

This study explores a multi-modal machine-learning-based approach to classify solitary pulmonary nodules (SPNs). Non-small cell lung cancer (NSCLC), presenting primarily as SPNs, is the leading cause of cancer-related deaths worldwide. Early detection and appropriate management of SPNs are critical to improving patient outcomes, necessitating efficient diagnostic methodologies. While CT and PET scans are pivotal in the diagnostic process, their interpretation remains prone to human error and delays in treatment implementation. This study proposes a machine-learning-based network to mitigate these concerns, integrating CT, PET, and manually extracted features in a multi-modal manner by integrating multiple image modalities and tabular features). CT and PET images are classified by a VGG19 network, while additional SPN features in combination with the outputs of VGG19 are processed by an XGBoost model to perform the ultimate diagnosis. The proposed methodology is evaluated using patient data from the Department of Nuclear Medicine of the University Hospital of Patras in Greece. We used 402 patient cases with human annotations to internally validate the model and 96 histopathological-confirmed cases for external evaluation. The model exhibited 97% agreement with the human readers and 85% diagnostic performance in the external set. It also identified the VGG19 predictions from CT and PET images, SUVmax, and diameter as key malignancy predictors. The study suggests that combining all available image modalities and SPN characteristics improves the agreement of the model with the human readers and the diagnostic efficiency. [ABSTRACT FROM AUTHOR]

This paper addresses the challenges of calibrating low-cost electrochemical sensor systems for air quality monitoring. The proliferation of pollutants in the atmosphere necessitates efficient monitoring systems, and low-cost sensors offer a promising solution. However, issues such as drift, cross-sensitivity, and inter-unit consistency have raised concerns about their accuracy and reliability. The study explores the following three calibration methods for converting sensor signals to concentration measurements: utilizing manufacturer-provided equations, incorporating machine learning (ML) algorithms, and directly applying ML to voltage signals. Experiments were performed in three urban sites in Greece. High-end instrumentation provided the reference concentrations for training and evaluation of the model. The results reveal that utilizing voltage signals instead of the manufacturer's calibration equations diminishes variability among identical sensors. Moreover, the latter approach enhances calibration efficiency for CO, NO, NO2, and O3 sensors while incorporating voltage signals from all sensors in the ML algorithm, taking advantage of cross-sensitivity to improve calibration performance. The Random Forest ML algorithm is a promising solution for calibrating similar devices for use in urban areas. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

The SmartAQ (Smart Air Quality) forecasting system produces high-resolution (1 × 1 km2) air quality predictions in an urban area for the next three days using advanced chemical transport modeling. In this study, we evaluated the SmartAQ performance for the urban area of Patras, Greece, for four months (July 2021, September 2021, December 2021, and March 2022), covering all seasons. In this work, we assess the system's ability to forecast PM2.5 levels and the major gas-phase pollutants during periods with different meteorological conditions and local emissions, but also in areas of the city with different characteristics (urban, suburban, and background sites). We take advantage of this SmartAQ application to also quantify the main sources of the pollutants at each site. During the summertime, PM2.5 model performance was excellent (Fbias < 15%, Ferror < 30%) for all sites both in the city center and suburbs. For the city center, the model reproduced well (MB = −0.9 μg m−3, ME = 2.5 μg m−3) the overall measured PM2.5 behavior and the high nighttime peaks due to cooking activity, as well as the transported PM pollution in the suburbs. During the fall, the SmartAQ PM2.5 performance was good (Fbias < 42%, Ferror < 45%) for the city center and the suburban core, while it was average (Fbias < 50%, Ferror < 54%, MB, ME < 3.3 μg m−3) for the suburbs because the model overpredicted the long-range transport of pollution. For wintertime, the system reproduced well (MB = −2 μg m−3, ME = 6.5 μg m−3) the PM2.5 concentration in the high-biomass-burning emission area with an excellent model performance (Fbias = −4%, Ferror = 33%) and reproduced well (MB < 1.1 μg m−3, ME < 3 μg m−3) the background PM2.5 levels. SmartAQ reproduced well the PM2.5 concentrations in the urban and suburban core during the spring (Fbias < 40%, Ferror < 50%, MB < 8.5 μg m−3, ME < 10 μg m−3), while it tended to slightly overestimate the regional pollution. The main local source of fine PM during summer and autumn was cooking, but most of the PM was transported to the city. Residential biomass burning was the dominant particle source of pollution during winter and early spring. For gas-phase pollutants, the system reproduced well the daily nitrogen oxides (NOx) concentrations during the summertime. Predicted NOx concentrations during the winter were consistent with measurements at night but underestimated the observations during the rest of the day. SmartAQ achieved the US EPA modeling goals for hourly O3 concentrations indicating good model performance. [ABSTRACT FROM AUTHOR]

Myocardial Perfusion Imaging (MPI) has played a central role in the non-invasive identification of patients with Coronary Artery Disease (CAD). Clinical factors, such as recurrent diseases, predisposing factors, and diagnostic tests, also play a vital role. However, none of these factors offer a straightforward and reliable indication, making the diagnosis of CAD a non-trivial task for nuclear medicine experts. While Machine Learning (ML) and Deep Learning (DL) techniques have shown promise in this domain, their "black-box" nature remains a significant barrier to clinical adoption, a challenge that the existing literature has not yet fully addressed. This study introduces the Deep Fuzzy Cognitive Map (DeepFCM), a novel, transparent, and explainable model designed to diagnose CAD using imaging and clinical data. DeepFCM employs an inner Convolutional Neural Network (CNN) to classify MPI polar map images. The CNN's prediction is combined with clinical data by the FCM-based classifier to reach an outcome regarding the presence of CAD. For the initialization of interconnections among DeepFCM concepts, expert knowledge is provided. Particle Swarm Optimization (PSO) is utilized to adjust the weight values to the correlated dataset and expert knowledge. The model's key advantage lies in its explainability, provided through three main functionalities. First, DeepFCM integrates a Gradient Class Activation Mapping (Grad-CAM) algorithm to highlight significant regions on the polar maps. Second, DeepFCM discloses its internal weights and their impact on the diagnostic outcome. Third, the model employs the Generative Pre-trained Transformer (GPT) version 3.5 model to generate meaningful explanations for medical staff. Our dataset comprises 594 patients, who underwent invasive coronary angiography (ICA) at the department of Nuclear Medicine of the University Hospital of Patras in Greece. As far as the classification results are concerned, DeepFCM achieved an accuracy of 83.07%, a sensitivity of 86.21%, and a specificity of 79.99%. The explainability-enhancing methods were assessed by the medical experts on the authors' team and are presented within. The proposed framework can have immediate application in daily routines and can also serve educational purposes. [ABSTRACT FROM AUTHOR]

The pre-operative localisation of abnormal parathyroid glands (PG) in parathyroid scintigraphy is essential for suggesting treatment and assisting surgery. Human experts examine the scintigraphic image outputs. An assisting diagnostic framework for localisation reduces the workload of physicians and can serve educational purposes. Former studies from the authors suggested a successful deep learning model, but it produced many false positives. Between 2010 and 2020, 648 participants were enrolled in the Department of Nuclear Medicine of the University Hospital of Patras, Greece. An innovative modification of the well-known VGG19 network (ParaNet+) is proposed to classify scintigraphic images into normal and abnormal classes. The Grad-CAM++ algorithm is applied to localise the abnormal PGs. An external dataset of 100 patients imaged at the same department who underwent parathyroidectomy in 2021 and 2022 was used for evaluation. ParaNet+ agreed with the human readers, showing 0.9861 on a patient-level and 0.8831 on a PG-level basis under a 10-fold cross-validation on the training set of 648 participants. Regarding the external dataset, the experts identified 93 of 100 abnormal patient cases and 99 of 118 surgically excised abnormal PGs. The human-reader false-positive rate (FPR) was 10% on a PG basis. ParaNet+ identified 99/100 abnormal cases and 103/118 PGs, with an 11.2% FPR. The model achieved higher sensitivity on both patient and PG bases than the human reader (99.0% vs. 93% and 87.3% vs. 83.9%, respectively), with comparable FPRs. Deep learning can assist in detecting and localising abnormal PGs in scintigraphic scans of patients with primary hyperparathyroidism and can be adapted to the everyday routine. [ABSTRACT FROM AUTHOR]

Featured Application: An explainable computer-aided diagnosis system for cardiovascular diseases can be a valuable tool for primary health care. Medical experts can utilize such tools to pinpoint unhealthy patients accurately and early, hence decongesting the National Healthcare Service (NHS). In recent times, coronary artery disease (CAD) prediction and diagnosis have been the subject of many Medical decision support systems (MDSS) that make use of machine learning (ML) and deep learning (DL) algorithms. The common ground of most of these applications is that they function as black boxes. They reach a conclusion/diagnosis using multiple features as input; however, the user is oftentimes oblivious to the prediction process and the feature weights leading to the eventual prediction. The primary objective of this study is to enhance the transparency and comprehensibility of a black-box prediction model designed for CAD. The dataset employed in this research comprises biometric and clinical information obtained from 571 patients, encompassing 21 different features. Among the instances, 43% of cases of CAD were confirmed through invasive coronary angiography (ICA). Furthermore, a prediction model utilizing the aforementioned dataset and the CatBoost algorithm is analyzed to highlight its prediction making process and the significance of each input datum. State-of-the-art explainability mechanics are employed to highlight the significance of each feature, and common patterns and differences with the medical bibliography are then discussed. Moreover, the findings are compared with common risk factors for CAD, to offer an evaluation of the prediction process from the medical expert's point of view. By depicting how the algorithm weights the information contained in features, we shed light on the black-box mechanics of ML prediction models; by analyzing the findings, we explore their validity in accordance with the medical literature on the matter. [ABSTRACT FROM AUTHOR]

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 99mTc-sestamibi (MIBI). This methodology could be applied to the everyday routine of medics for real-time evaluation or educational purposes. [ABSTRACT FROM AUTHOR]

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.