Journal of Medical Engineering and Technology最新文献

Background: Recently, dynamic total-body positron emission tomography (PET) imaging has become possible due to new scanner devices. However, there is still little research systematically evaluating clustering algorithms for processing of dynamic total-body PET images.

Materials and methods: Here, we compare the performance of 15 unsupervised clustering methods, including K-means either by itself or after principal component analysis (PCA) or independent component analysis (ICA), Gaussian mixture model (GMM), fuzzy c-means (FCM), agglomerative clustering, spectral clustering, and several newer clustering algorithms, for classifying time activity curves (TACs) in dynamic PET images. We use dynamic total-body ¹⁵O-water PET images of 30 patients. To evaluate the clustering algorithms in a quantitative way, we use them to classify 5000 TACs from each image based on whether the curve is taken from brain, right heart ventricle, right kidney, lower right lung lobe, or urinary bladder.

Results: According to our results, the best methods are GMM, FCM, and ICA combined with mini batch K-means, which classified the TACs with a median accuracies of 89%, 83%, and 81%, respectively, in a processing time of half a second or less.

Conclusion: GMM, FCM, and ICA with mini batch K-means show promise for dynamic total-body PET analysis.

Cardiovascular diseases (CVDs) significantly impact athletes, impacting the heart and blood vessels. This article introduces a novel method to assess CVD in athletes through an artificial neural network (ANN). The model utilises the mutual learning-based artificial bee colony (ML-ABC) algorithm to set initial weights and proximal policy optimisation (PPO) to address imbalanced classification. ML-ABC uses mutual learning to enhance the learning process by updating the positions of the food sources with respect to the best fitness outcomes of two randomly selected individuals. PPO makes updates in the ANN stable and efficient to improve the model's reliability. Our approach formulates the classification problem as a series of decision-making processes, rewarding every classification act with higher rewards for correctly identifying the instances of the minority class, hence handling class imbalance. We evaluated the model's performance on a diversified medical dataset including 26,002 athletes who were examined within the Polyclinic for Occupational Health and Sports in Zagreb, further validated with NCAA and NHANES datasets to verify generalisability. Our findings indicate that our model outperforms existing models with accuracies of 0.88, 0.86 and 0.82 for the respective datasets. These results enhance clinical model application and advance cardiovascular disorder detection and methodologies.

Uroflowmetry plays an important role in the investigation of patients with lower urinary tract symptoms. We were required to assess a newly developed uroflowmeter. We thus aimed to produce a standardised protocol to test the accuracy and filtering of any new uroflowmeter. The accuracy of a newly developed uroflowmeter (Minze Uroflow^®) was validated using a constant flow bottle and a cylindrical flow column. Two other machines were also tested alongside. We also assessed filtering by reproducing common artefacts in the laboratory. Finally, a questionnaire was constructed to assess the usability of the uroflowmeter by clinicians during a normal hospital flow clinic. A protocol to test new uroflowmeters was written and assessed. The protocol showed the following results for the tested uroflowmeters: a simple bench test using a constant flow bottle and cylindric column showed that the uroflow parameters (Q_max and V_void) were within the claimed accuracy range and ICS recommendations. The processing of the flow data by the systems effectively filtered noise, and the flow rate decline over the whole measurement range, as produced by the cylindrical flow column, was smooth and linear. Usability was assessed by clinicians in their routine clinical practice. The proposed tests meet the requirements of the ICS guidelines. We have designed a protocol which can be used by clinicians and researchers to validate the accuracy of their uroflowmeters, evaluate new models and ensure clinical usefulness.

This review article delves into the cutting-edge realm of 3D printing and its impact on the fabrication of customised orthodontic retainers, which is an essential utility in the prevention of relapse post orthodontic treatment. This review evaluates the use of biocompatible materials and provides insight into future perspectives and improvements in this field. It highlights the potential of data collecting method and 3D printing to improve orthodontic retainers' fabrication and emphasises the importance of using biocompatible materials for patient safety and efficacy. It also explains cytotoxic qualities of retainer fabrication materials, which are vital for safeguarding the oral health of the patient. The evaluation procedure enables the early diagnosis and correction of any potential difficulties, such as maladjustment or inappropriate fit, allowing for a more effective treatment. It illustrates the breakthroughs and innovations in the field of orthodontics, the advantages of 3D printing over conventional methods, as well as the advantages and disadvantages of various fabrication method. Incorporating 3D printing and review into the production of orthodontic retainers enhances the overall effectiveness and efficiency of patient treatment.

Our work presents a real-time embedded implementation of a proposed approach for physiological signs monitoring, such as heart and breathing rates, using a Photoplethysmography signal (PPG) retrieved from digital RGB cameras. The proposed algorithm was implemented in an embedded architecture to assess both the processing time and algorithmic complexity. The proposed method is based on image processing techniques to extract the noisy PPG signal and signal processing, filtering, and decomposition algorithm to estimate the instantaneous vitals indicators. On the embedded implementation side, the common criteria that must be studied are the accuracy of the result estimation, processing time optimisation, and hardware-software adoption. The latter standard is met by the hardware-software co-design concept which will lead to adopting the algorithm's layers with an embedded platform architecture. On our side, we will principally use the High-Level Synthesis (HLS) as a parallel programming language and the computing homogeneous/heterogeneous devices (CPU/GPU). Our proposed optimised algorithm's implementation offers a gain of x5.05, x24.96, and x36.68 compared with the native version using MATLAB and the optimised version using C/C++, OpenMP, and OpenCL tool, respectively, in some functional blocks.

Present work involves rigorous experimentation for classification of mammographic masses by employing four deep transfer learning models using hierarchical framework. Experimental work is carried on 518 SFM images of DDSM dataset with 208, 150 and 160 images of probably benign, suspicious- malignant and highly malignant classes, respectively. ResNet50 model is used for generating segmented mass images. For hierarchical classification framework, at node 1, the segmented mass image is classified as belonging to probably benign (BIRAD-3) class or suspicious abnormality (BIRAD-4 and BIRAD-5) class. At node 2, the segmented mass image belonging to suspicious abnormality class is further classified as suspicious malignant (BIRAD-4) class or highly malignant (BIRAD-5) class. Deep transfer learning based hierarchical CAD systems experimented in the present work include VGG16/VGG19/ GoogleNet/ResNet50 models. It was noted that deep transfer learning model VGG19 at node 1 and VGG16 at node 2, yielded highest classification accuracy of 93 % and 90 %, respectively, therefore, a deep transfer learning based hybrid hierarchical CAD system was developed by employing VGG19 at node 1 and VGG16 at node 2. This model yields overall classification accuracy of 88 %. Further, hybrid hierarchical CAD system was designed using VGG19/ANFC-LH classifier at node 1, and VGG16/ANFC-LH classifier at node 2 yielding the highest classification accuracy of 92%. The promising result yielded by hybrid hierarchical CAD system design indicates its usefulness for step-wise classification of mammographic masses.