Volume 70, Issue 6

Lung sounds, as an important physiological signal of human body, play a crucial role in the diagnosis and monitoring of respiratory diseases. In recent years, deep learning-based lung sound intelligent recognition technology has made significant progress in the field of medical auxiliary diagnosis, particularly with the deployment of deep learning models on edge devices to achieve local inference and low-latency diagnosis. This has become a key direction in the development of lung sound recognition systems. This paper systematically reviews the fundamental processes of lung sound recognition and analysis, focusing on the construction of lung sound classification models, model lightweight design, and the latest research progress in deploying these models on embedded systems, FPGA, and other hardware platforms. Additionally, the paper looks forward to the application prospects and development trends of this technology in the field of intelligent lung sound recognition. In particular, deployment pathways grounded in hardware–software co-design on embedded and edge platforms are poised to further advance health monitoring systems. This paper provides a comprehensive reference for understanding the key aspects of lung sound recognition technology, from algorithm research to hardware implementation.

Objectives Accurate and timely swallowing ability screening is essential for effective prevention and treatment of dysphagia. Current clinical imaging methods are often radioactive or invasive, limiting their applicability for routine monitoring. This study aims to develop a smartphone-based intelligent system for real-time screening of swallowing ability using a wearable throat microphone. Methods First, a customized Android application was developed to collect, visualize, and analyze sounds related to swallowing from a wearable throat microphone. Second, a transfer learning model, YAMNet-S, was trained on 4,715 one-second audio segments of coughing, swallowing, and other noises, obtained from 215 healthy participants (Experiment 1). The trained model was then deployed on a smartphone to classify swallowing-related events in real time. Finally, the water swallow test (WST) was conducted by counting swallowing and coughing events within 30 s from 15 simulated patients with self-induced voluntary coughs for clinical validation (Experiment 2). Results The mean accuracy of the trained model in swallowing and coughing events classification is 94.48 and 94.45 %, respectively, and the difference of WST scores in system calculations and expert evaluation was 0.267 (out of 5). Conclusions This smartphone-based intelligent system has the potential to be a comfort, portability and user-friendliness tool for preliminary screening of dysphagia before VFSS/FEES, especially on some situations with limited medical resources, such as community or home-based care.

Objectives Flight cognitive ability is the core element of guaranteeing flight safety. This study attempts to explore the basic cognitive operation ability of pilots under overload environment, and provides reference for the research on the index screening and monitoring technology methods cognitive operation ability under centrifuge overload condition. Methods A specific cognitive operation ability task is designed, physiological signals such as electroencephalography (EEG) when performing cognitive ability tasks under human centrifuge overload condition are obtained, and the characteristic parameters changes are studied. Results When performing cognitive tasks, both the beta waves and the Theta/Beta Ratio (TBR) parameters in EEG show certain specific changes after the task starts and in the later stage the task. Conclusions Beta wave energy ratio and TBR parameter can better reflect the changes of pilot’s cognition under overload environment, and the joint research of multiple frequency bands and multiple parameters helps to find the parameters and indicators with the potential to become the physiological monitoring index of pilot’s cognitive operation capacity under overload condition more quickly and better. This study lays a foundation and provides a reference for the follow-up selection of physiological monitoring index of a pilot’s cognitive operation ability and the research of monitoring technology on cognitive operation ability under overload condition.

Objectives This paper investigates an efficient UWB biosensor that employs monostatic radar-based microwave imaging (RBMI) to identify cancerous cells in human breast tissue. Methods The proposed FSS-loaded lens applicator biosensor assembly combines three designs, including a UWB radiator, a frequency selective surface (FSS) reflector, and a square-shaped split ring metamaterial cell (SSR-M). In Computer Simulation Technology-Microwave Studio (CST-MWS) simulator, a lens applicator (UWB transceiver with SSR-M at its front side) when loaded with an FSS reflector is simulated with a breast phantom (BP), then the system performance is assessed in the context of backscattered parameters, penetration depth (PD), and specific absorption rate (SAR). Experimental validations of the fabricated prototype are captured using the Vector Network Analyzer (VNA). Results/Key findings Signal processing steps are applied over the S 11 signals using MATLAB, to detect the depth/location of the malignant tissue. The proposed system is capable of detecting a minimum tumor size of 4 mm up to a maximum localization depth of 30 mm. The proposed system has an isotropic gain of 3.6 dBi and a SAR of 1.07 W/kg over 1 g of tissue. Conclusions As a result, the proposed biosensor configuration is suitable option for microwave imaging applications.

Objectives Small animal models, particularly mice, are crucial for studying gastrointestinal diseases like colorectal cancer. Tumor assessment via colonoscopy generates large video datasets, necessitating automated analysis due to limited resources and time-consuming manual review. Methods We employed a YOLOv7-based deep learning model pre-trained on human polyp images to detect tumors in mouse colonoscopy videos. Detection was enhanced using a stool detector and a color-based filter. Lesions were classified from ‘0’ (no tumor) to ‘5’ (tumor >50 % of colon diameter) using a custom ratio-based method. The system was evaluated on 150 videos from 28 mice over 6 weeks, with 125 videos containing tumors. Results Initial detection yielded a Precision of 0.576, Recall of 0.916, and Accuracy of 0.593. Adding the stool detector improved results to 0.932, 0.946, and 0.897, respectively. Compared to expert annotations, classification reached 0.759 Precision, 0.774 Recall, and 0.774 Accuracy over all five classes. Conclusions The proposed approach reliably detects and classifies colon tumors in mice, offering real-time support for preclinical endoscopic studies. Further evaluation will provide more insights into its performance.

Objectives Accurate segmentation of the temporomandibular joint (TMJ) in cone beam computed tomography (CBCT) images is crucial for the diagnosis of temporomandibular joint osteoarthritis (TMJ-OA), particularly of the condyle. The aim of this study was to use a multi-stage image pyramid network for precise segmentation of the TMJ in CBCT images. Methods A multi-stage image pyramid network (MIP-Net) was proposed in this study for accurate TMJ segmentation in CBCT images. This network utilizes stepwise down-sampling to form an image pyramid, and feature fusion is employed to integrate features of different scales, resulting in accurate segmentation of the TMJ in CBCT images. Results The network was trained and evaluated using a clinical image dataset, and the dice similarity coefficient (DSC) was 98.02 ± 0.65 %, the average surface distance (ASD) was 0.0605 ± 0.0229 mm, and the Hausdorff distance (HD) was 0.1907 ± 0.0627 mm. Conclusions The proposed MIP-Net achieved accurate segmentation of the TMJ in CBCT images, which may help to improve the clinical diagnosis and treatment of TMJ-OA.

Objectives This study aims to optimize the design of porous dental implants using triply periodic minimal surfaces (TPMS) to address stress shielding caused by elastic modulus mismatch between titanium implants and bone tissue, while enhancing osseointegration through controlled porosity. Methods Two TPMS architectures (D-type and G-type) were modeled via MathMod and Rhino software, with porosity controlled by parameter t. Finite element analysis (FEA) evaluated mechanical properties under porosities of 40 %, 60 %, and 80 %, and stress distribution in a patient-specific mandibular model under 200 N masticatory load. Results Both D- and G-type structures met mechanical requirements at ≤60 % porosity. The D-type exhibited optimal stress transfer at 60 % porosity, whereas G-type at 80 % porosity exceeded yield strength due to thin walls. Porous designs reduced cortical-cancellous bone stress differences by 30–50 %, mitigating stress shielding. The interconnected network facilitated bone ingrowth. Conclusions TPMS-based porous implants with 40–60 % porosity balance mechanical strength and stress distribution, offering personalized solutions for bone quality adaptation. Future work should validate long-term stability through in vitro experiments.