Physical and Engineering Sciences in Medicine最新文献

Glioma is primarily treated through surgical resection, but accurately identifying tumor boundaries remains challenging. Traditional intraoperative diagnostic techniques, such as frozen section pathological examination and intraoperative magnetic resonance imaging, suffer from issues such as long duration, high cost, and complex operation. A rapid and accurate intraoperative auxiliary diagnostic method for glioma based on the differences in dielectric properties combined with machine learning is proposed in this study. Using an open-ended coaxial probe technique, the dielectric properties of 81 glioma tissue samples and 47 normal brain tissue samples from 14 patients were measured over a frequency range of 1 MHz-4 GHz. After feature selection and dimensionality reduction using the Lasso method, four machine learning models-Naive Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Artificial Neural Network (ANN)-were used to classify the samples. Model performance was evaluated using accuracy, precision, recall, F1 score, and the area under the Receiver Operating Characteristic curve (AUC value). The experimental results demonstrated that the dielectric properties of glioma tissues are higher than those of normal brain tissues (with an average increase of 22% in conductivity and 18% in relative permittivity). On the test set, the KNN model exhibited the highest classification accuracy (90%), while the ANN model showed the best AUC value (0.95). This study confirms that the rapid identification of glioma can be achieved based on dielectric properties combined with machine learning techniques, providing neurosurgeons with a novel auxiliary diagnostic technology for precise intraoperative margin detection of glioma.

A prototype Compton camera composed of two high resolution scintillator detectors is presented in this work. The scatterer detector consists of a 21 × 21 gadolinium aluminum gallium garnet (GAGG) crystal array with a crystal size of 0.6 × 0.6 × 2 mm³. The absorber detector consists of a 23 × 23 lutetium yttrium orthosilicate (LYSO) crystal array with a crystal size of 1.0 × 1.0 × 20 mm³. A simple back-projection image reconstruction method was developed. The energy of the scatterer detector was accurately calibrated using the 55, 202, 307 keV gamma-rays from the LYSO natural background and the 511 keV gamma-ray from a ²²Na point source. The scatterer detector provides a performance with all crystals clearly resolved even at an energy window of 30-120 keV and an average crystal energy resolution of 10.4% at 511 keV. The absorber detector provides a performance with all crystals clearly resolved, an average crystal depth of interaction resolution of ~ 2 mm and an average crystal energy resolution of 19.4% at 511 keV. An average spatial resolution of 2.5 mm was obtained and 9 point sources of 3 mm apart were well resolved at an image plane 7.5 mm from the front of the scatterer detector by using the 511 keV gamma-rays from a ²²Na point sources. Furthermore, iterative reconstruction using the maximum-likelihood expectation maximization (MLEM) algorithm achieved a spatial resolution of ~ 1 mm at a plane 7.5 mm from the front of the scatterer detector. Compared with the simple back-projection method, the MLEM reconstruction significantly enhanced the image contrast and effectively suppressed the background artifacts.

Separate renal function assessment is important in clinical decision making. The single-photon emission computed tomography is commonly used for the assessment although radioactive, tedious and of high cost. This study aimed to automatically assess the separate renal function using plain CT images and artificial intelligence methods, including deep learning-based automatic segmentation and radiomics modeling. We performed a retrospective study on 281 patients with nephrarctia or hydronephrosis from two centers (Training set: 159 patients from Center I; Test set: 122 patients from Center II). The renal parenchyma and hydronephrosis regions in plain CT images were automatically segmented using deep learning-based U-Net transformers (UNETR). Radiomic features were extracted from the two regions and used to build radiomic signature using the ElasticNet, then further combined with clinical characteristics using multivariable logistic regression to obtain an integrated model. The automatic segmentation was evaluated using the dice similarity coefficient (DSC). The mean DSC of automatic kidney segmentation based on UNETR was 0.894 and 0.881 in the training and test sets. The average time of automatic and manual segmentation was 3.4 s/case and 1477.9 s/case. The AUC of radiomic signature was 0.778 in the training set and 0.801 in the test set. The AUC of the integrated model was 0.792 and 0.825 in the training and test sets. It is feasible to assess the renal function of each kidney separately using plain CT and AI methods. Our method can minimize the radiation risk, improve the diagnostic efficiency and reduce the costs.

Work-related musculoskeletal disorders represent a significant occupational health issue. These disorders encompass a range of conditions resulting from specific risk factors associate to manual material handling such as: intensity, repetition, and duration. Over the years, several observational methodologies have been developed to assess biomechanical risk, but their limits depend mainly on clinicians' subjective assessment. For this reason, wearable sensors coupled with artificial intelligence have recently been integrated in the occupational ergonomic field. This study aimed to develop a new technological methodology-based on machine learning algorithms and inertial wearable sensors-able to automatically discriminate biomechanical risk associated with lifting loads. Ten healthy volunteers were enrolled in this study performing specific weight-lifting tasks wearing two inertial measurement units on the sternum and lumbar region. The acquired inertial signals were appropriately processed to extract several features in the time-domain and frequency-domain which have been used as input to several machine learning algorithms. Excellent results in discriminating biomechanical risk classes were obtained reaching accuracies and areas under the receiver operating characteristic curve above 86% and 95%, respectively. In addition, the sternum emerged as the most informative body landmark, while the mean absolute value was identified as the most informative feature. Future investigations on a larger study population could confirm the potential of the proposed automatic procedure to be used in the workplace in combination with well-established methodologies.

Accurate differentiation between non-cancerous, benign, and malignant lung cancer remains a diagnostic challenge due to overlapping clinical and imaging characteristics. This study proposes a multimodal machine learning (ML) framework integrating positron emission tomography/computed tomography (PET/CT) anatomic-metabolic parameters, sarcopenia markers, and inflammatory biomarkers to enhance classification performance in lung cancer. A retrospective dataset of 222 patients was analyzed, including demographic variables, functional and morphometric sarcopenia indices, hematological inflammation markers, and PET/CT derived parameters such as maximum and mean standardized uptake value (SUVmax, SUVmean), metabolic tumor volume (MTV), total lesion glycolysis (TLG). Five ML algorithms-Logistic Regression, Multi-Layer Perceptron, Support Vector Machine, Extreme Gradient Boosting, and Random Forest-were evaluated using standardized performance metrics. Synthetic Minority Oversampling Technique was applied to balance class distributions. Feature importance analysis was conducted using the optimal model, and classification was repeated using the top 15 features. Among the models, Random Forest demonstrated superior predictive performance with a test accuracy of 96%, precision, recall, and F1-score of 0.96, and an average AUC of 0.99. Feature importance analysis revealed SUVmax, SUVmean, total lesion glycolysis, and skeletal muscle index as leading predictors. A secondary classification using only the top 15 features yielded even higher test accuracy (97%). These findings underscore the potential of integrating metabolic imaging, physical function, and biochemical inflammation markers in a non-invasive ML-based diagnostic pipeline. The proposed framework demonstrates high accuracy and generalizability and may serve as an effective clinical decision support tool in early lung cancer diagnosis and risk stratification.

Dual-energy computed tomography (DECT) generates virtual monochromatic images (VMI) and material decomposition images (MDI), facilitating enhanced tissue contrast and quantitative material assessment. However, the accuracy of these measurements may be influenced by object size due to beam hardening and associated spectral changes. To evaluate the impact of object size on the accuracy of iodine quantification and CT numbers in virtual monochromatic images (VMI) using split-filter dual-energy CT (SFDE), and to compare its performance with sequential acquisition dual-energy CT (SADE). CT scans were performed on phantoms with diameters ranging from 16 to 36 cm using both SFDE and SADE techniques. Virtual monochromatic images and material decomposition images were generated. CT numbers and iodine concentrations were measured from embedded iodine rods, and relative errors were calculated using the 16 cm phantom as a reference. CT numbers in VMI obtained from SFDE exhibited increasing variability with larger phantom sizes, particularly at both low and high energy levels. Iodine quantification errors with SFDE exceeded 10% in all phantom sizes and reached approximately 60% in the 36 cm phantom. In contrast, SADE consistently maintained measurement errors within 10%. Object size significantly influences the accuracy of CT numbers and iodine quantification using SFDE, with larger phantoms showing marked overestimation. These results suggest that careful interpretation is necessary when applying SFDE-based quantitative imaging in patients with larger object sizes.

Seismocardiography (SCG) uses accelerometers to record cardiac-induced accelerations of the chest wall. Cardiorespiratory interactions cause changes in amplitude and morphology of the SCG signals. Accelerometers can also directly monitor respiration by tracking thoracic inclination. This study thoroughly investigated the influence of accelerometer placement on the monitoring accuracy of respiration and cardiorespiratory interactions from SCG signals. Simultaneous recordings acquired by 16 accelerometers and a respiration belt placed onto 9 subjects' chests were analyzed. Respiratory signals were estimated considering: (a) chest inclination, (b) amplitude modulation (AM) and (c) morphological changes of SCG signals for each sensor location. For the first time in literature, a continuous description of respiratory-induced changes in SCG morphology was obtained via a morphological similarity index (MSi). The performance of respiratory acts detection and inter-breath intervals (IBIs) estimation was evaluated against the concurrent reference respiration signal. High accuracy was achieved in all three kinds of respiratory signals, with average sensitivity and positive predictive value of 95.8% and 95.5% for chest inclination, 85.9% and 84.4% for AM, 94.3% and 95.7% for MSi. Moreover, IBIs measurements showed non-significant biases and limits of agreement of about ± 0.8 s for chest inclination and MSi, and ± 1 s for AM. Performance achieved by chest inclination and MSi appeared not much influenced by sensor location, while AM showed higher variations. Information on breathing and cardiorespiratory interactions can be accurately obtained via SCG on multiple sites on the chest.

The study of dielectric properties (DPs) reveals significant differences between normal and liver cancer tissues. Although the open-ended coaxial probe (OCP) method is widely used for measuring DPs, tumor infiltration depth affects the measurements, blurring dielectric thresholds and posing challenges for tissue identification based on DPs. This study combines DPs analysis with machine learning (ML) to achieve two key goals: (1) accurately distinguish tissue types, (2) reliably predict tumor infiltration depth. We simulated the DPs of liver cancer tissues at different infiltration depths, using a total of 90,000 samples with 181 frequency-point features. We evaluated the performance of common ML models, including artificial neural networks (ANN), support vector machines (SVM), and Bagging tree ensembles, and validated them using real tissue and phantom measurements. Additionally, the probe's detection depth was experimentally validated. Experimental results showed that all three ML models performed well in tissue identification and tumor infiltration depth prediction. SVM achieved the highest classification accuracy of 98.91%. For depth prediction, SVM and ANN yielded MAPE/RMSE of 0.1742/0.0673 and 0.1658/0.0730, respectively. The probe's effective detection range was 0.1-0.6 mm, essential for accurate measurement and prediction. The models also demonstrated strong performance in real tissue and phantom validations, with the Bagging ensemble achieving 100% classification accuracy and MAPE/RMSE of 0.1434/0.0614 for prediction. These findings confirm the method's reliability for precise tissue identification and infiltration depth estimation, supporting accurate tumor resection and improved patient outcomes.

Accurate prediction of surface doses is crucial for clinical outcomes in radiotherapy. Surface dose distribution must be predicted accurately by calculation algorithms in the treatment planning system (TPS). This study aims to compare surface dose calculations from the Eclipse TPS with radiochromic film measurements to evaluate the reliability of these calculation algorithms. Measurements with radiochromic films were performed using 6 MV photon beams. Treatment plans for 3D conformal radiotherapy (3DCRT), intensity-modulated radiotherapy (IMRT), and volumetric arc therapy (VMAT) were generated on the TPS and calculated using various algorithms. Treatment plans were irradiated on Gafchromic EBT3 films with a PTW head and neck phantom. EBT3 films were compared to calculation algorithms via FilmQA™ Pro (version 7.0) software with multi-channel analysis. Dosimetric evaluations were statistically analyzed. Commercial calculation algorithms underestimated the surface dose in 3DCRT, IMRT, and VMAT treatment plans. For 3DCRT, the underestimations were 8.0% with the AAA algorithm and 8.7% with AXB. In VMAT, the underestimations were 10.2% with AAA and 12.9% with AXB. For IMRT, the underestimations were 6.6% with AAA and 7.3% with AXB. The AAA algorithm closely matched surface dose measurements among calculation methods. The dosimetric results indicate that both AAA and AXB algorithms, as implemented in the Eclipse™ TPS, tend to underestimate surface dose compared to EBT3 film measurements. Accurate knowledge of the dose in the superficial region is crucial to prevent acute skin reactions or to deliver an effective dose to superficial tumors in clinically significant cases. Therefore, our surface dose measurements offer more accurate evaluations, making Gafchromic EBT3 films suitable for such cases.

The World Health Organisation 2024 report shows that Cardiovascular Disease (CVD) is the leading cause of death worldwide, estimated at 17.9 million deaths annually, and its mortality is about 32% of all deaths in the world. Of these, about 85% are myocardial infarctions and strokes. This study aims to diagnose heart disorders by providing early medical intervention to reduce the risks of abnormal heart structures. A data-driven model has been developed to achieve the above aim. The CVD and standard Electrocardiogram (ECG) datasets are extracted from PhysioNet in CSV format. This dataset comprises 305 samples of normal heart function, 15 samples of congestive heart failure, 32 samples of intracardiac atrial fibrillation, and 77 samples of supraventricular arrhythmia. The key steps include preprocessing the raw ECG data, extracting the relevant features, and introducing the input to the Machine Learning (ML) model for training. After preprocessing, ECG characteristic features, viz., mean heart interval, RR interval, p-wave amplitude, q-wave amplitude, r-wave amplitude, t-wave amplitude, and the derived features, namely, root mean square of successive difference (RMSSD), mean standard deviation of the normal-to-normal interval (SDDN), are extracted from the ECG signal and implemented using eXplainable Artificial Intelligence (XAI) methods to expound feature contributions. Various ML algorithms, including ensemble (EN), Naive Bayes (NB), and Support Vector Machine (SVM), are implemented for effectiveness. A tenfold cross-validation and performance are assessed using accuracy and recall analysis. Among these four models, SVM outperforms the other models and feature selection, achieving 99.5% accuracy when considering all features, 77% accuracy for the two derived features, and 99.5% accuracy for ECG wave characteristics features. To address the limitations, such as a small dataset and class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) is applied to further enhance model performance. This study demonstrates the effectiveness of ML models, notably SVM, in predicting CVD abnormalities based on their ECG characteristics. These results suggest that future research should focus on refining methods to identify key features of ECG wave characteristics, potentially streamlining and speeding up the prediction of CVD in real-time. This work utilises XAI techniques to make the models more transparent, understandable and improve model accuracy of 99.8% for SVM. Furthermore, increasing model transparency with XAI might facilitate quicker clinical adoption for the diagnosis of heart disease.