Objective To explore the application effect of artifact removal technique of Discovery RT (GE Inc) in organ segmentation and radiotherapy planning. Methods Twelve patients with head and neck tumors carrying irremovable metal dentures,who received radiotherapy at the Fourth Affiliated Hospital of Nanjing Medical University from September 2021 to February 2024,were enrolled in the study. AccuContour was used to perform organs-at-risk (OAR) segmentations separately on CT image processed by metal artifact reducing algorithm and conventional CT image. Dice similarity coefficient (DSC),Hausdorff distance (HD) and the mean CT values of OAR with or without metal artifacts were compared. Both the processed CT images and original images were transferred to Monaco planning system. Dose calculation was conducted on the images with metal artifacts,and the calculated plans were mapped to the artifact-free images for analyzing the dosimetric differences of OAR. Results The HD of the throat,oral cavity and mandible were slightly higher. The mean CT values of the tissues in images before and after artifact removal were not statistically different except for the trachea (P<0.05),and the mean oral CT value at the artifact level differed significantly in two groups (P<0.05). The radiation dose of different tissues in the radiotherapy plan before and after artifact removal only showed statistical differences in the superior constrictor musde of pharynx. The percentages of irradiation dose differences for the tissues ranged from 0 to 6.45％,with the largest fluctuation in the left lens (-2.92％±3.95％) and the right lens (1.29％±3.35％). Conclusion Manual delineation is required for the tissues close to metal artifacts due to the remarkable effects on CT values and planned dose,but there is few differences in the other tissues.

Objective To synthesize high-quality synthetic CT (sCT) images from cone beam CT (CBCT) by learning lung CT domain image features with a deep learning algorithm. Methods A sCT generation algorithm which employs perceptual loss-based cyclic generative adversarial network model (CycleGAN) and iterative registration was presented. CycleGAN model was trained to generate high-quality sCT images by combining perceptual loss and cycle consistency loss;and Elastix was used to register the generated sCT image and the planned CT (pCT) image,and iterate CycleGAN generator model. Results Experiments were conducted on the obtained pCT and CBCT data of 70 patients with lung tumors. From a quantitative perspective,the SSIM between sCT generated by the proposed algorithm and pCT was improved by 11.9％ as compared with that between CBCT and pCT,increasing from 0.825 to 0.923;additionally,RMSE dropped from 110.97 HU to 78.62 HU,PSNR increased from 32.21 dB to 34.74 dB,and mutual information increased from 1.187 to 1.418. The visual evaluation revealed that the proposed algorithm greatly eliminated the scattering artifacts of CBCT slices,highlighted the bone structure,and repaired the soft tissue structure. The comparisons with U-CycleGAN,R-CycleGAN and CUT models confirmed the effectiveness of the proposed algorithm. Conclusion Using the proposed algorithm for sCT images generation can effectively reduce the dose error and structural error between CBCT and pCT,making it possible to apply the proposed algorithm to accurate dose calculations and assist doctors in clinical diagnosis.

The complex rhythmic patterns of central pattern generator (CPG) and its response to electromagnetic induction are explored. The response of an individual neuron's discharge activity to electromagnetic induction is analyzed when the slow variable parameter changes,and it is found that the smaller the slow variable parameter,the stronger the neuron's response to electromagnetic induction. The simulation reveals that the discharge rhythm of CPG that are sequentially inhibited (neuron 1 inhibits 2,neuron 2 inhibits 3,and neuron 3 inhibits 1) has a stronger response to electromagnetic induction in synchronous states and is more robust in typical discharge states (3 neurons firing in sequence). In addition,the response of the discharge rhythm of CPG with mutual inhibition coupling structure to electromagnetic induction is discussed,and it is found that small changes in electromagnetic induction will fundamentally transform the discharge rhythm pattern of CPG.

An epilepsy prediction model based on two-dimensional convolutional neural network and Cox-Stuart test for non-independent patients is proposed to address the problem of how to effectively predict whether epilepsy patients are going to have an attack or not. After EEG data normalization and EEG signal noise removal using a notch filter and a high-pass filter,the filtered signals are inputted into the two-dimensional convolutional neural network model for feature extraction and classification,and Cox-Stuart test is used to determine whether an early stopping is needed or not,thereby reducing the computational and time complexities of the model. The model is tested under the conditions with pre-seizure periods of 10,30 and 60 min,respectively,and the results show that the model performs best when the pre-seizure period is 10 min. The model has an average accuracy,sensitivity and specificity of 97.70％,97.36％and 98.04％on the test set,demonstrating its superior performance.

In response to the demand for portable and rapid measurement in normalized nutritional metabolism monitoring,a low-power and multifunctional measurement method and system scheme based on indirect calorimetry is proposed. Firstly,a combination of Kalman and the 5th order Bessel filtering algorithm is used to preprocess the real-time data of oxygen,carbon dioxide,flow rate and electrocardiogram,and the cumulative amount of gas per unit time is obtained. Secondly,a rapid nutrient metabolism measurement method that considers resting analyses of electrocardiogram and respiratory waves is designed,and a calculation model for energy metabolism parameters is constructed. Then,a dedicated collection and analysis circuit system is developed to transmit data to the upper computer or cloud through wireless networks,achieving intelligent and networked design. Finally,the resting state of the subjects is determined by analyzing electrocardiogram and respiratory parameters,and the test of monitoring nutrient metabolism is completed. The results show that the system can effectively detect metabolism related physiological signals with high precision,and achieve accurate measurement of human fat and sugar consumptions on metabolism,providing an effective means for measuring human metabolic energy.

Objective To design a new type of coil and use a shield plate to optimize the focusing area,so that the coil can meet the usage requirements to the greatest extent possible. Methods Based on a 3-layer head model,COMSOL simulation software was used to establish a real human head model. After effectiveness verification,the traditional 8-shaped coil,new coil,and head model were jointly simulated to explore the optimal performance parameters of the new coil. The focusing performance was further improved by adding a shield plate,and the effects of different shield plate window sizes on focusing performance were analyzed. Results Compared with the traditional 8-shaped coil,the new coil improved the magnetic field intensity in the brain by 32％,increased the stimulation depth by 152％,but slightly expanded the focusing area,decreasing the focusing performance by 7.2％,which demonstrated the superior performance of the new coil than the traditional 8-shaped coil. After optimization using a shield plate with a window size of 600 mm2,the focusing performance of the new coil was improved by 30％,verifying the effectiveness of the shield plate on the coil performance. Conclusion The feasibility of the new coil is proved in terms of performance parameters,and the effects of different coil sizes and shield plate sizes on evaluation indicators are explored,laying a foundation for subsequent research on transcranial magnetic stimulation.

Brain age prediction is of great significance to the in-depth understanding of individual neurodevelopment,early diagnosis of neuropsychiatric disorders,and formulation of personalized treatment plans. With the continuous advancement of deep learning,more and more researches focus on using such algorithms to predict brain age. Compared with traditional regression algorithms,deep learning which has the advantages of complex pattern learning,end-to-end learning and high adaptability can more accurately reveal the neuropathological mechanisms of neuropsychiatric disorders,and provide more precise tools for clinical assessment,assisted diagnosis and prognosis prediction. Herein the study reviews the recent advances in the application of deep learning algorithms in brain age prediction,introduces the achievements in deep learning model optimization,multimodal data inputs and interpretability studies for brain age prediction,discusses the methods for the establishment of integrated deep learning architectures and the future challenges of developing unified benchmarking,and provides an outlook on the application of deep learning in brain age prediction.

The construction of standardized and high-quality clinical research database is an important link for clinicians or researchers to carry out clinical medical research,and also an important guarantee for obtaining high-quality scientific research results and publication. A review on the types and challenges of traditional clinical research database creation is provided,focusing on the key technical solutions of constructing high-quality clinical research database based on electronic data capture (EDC) system,including 6 modules of technological innovation program:the combination of interactive web response system and EDC system,logical check,intelligent follow-up management,data dictionary,trace tracing,statistical analysis of research data,so as to solve the deficiency of traditional software system in constructing clinical research database. In addition,the applications of EDC system in clinical research database construction are summarized and its development prospects are discussed for providing references for clinicians or researchers to carry out high-quality clinical researches.

To address the challenges of detecting small nodules in lung CT images and achieving a balance between lightweight and high-precision with the existing lung nodule detection algorithms,a high-precision and lightweight lung nodule detection algorithm based on improved YOLOv5 is proposed. The main improvements are focused on 4 aspects. (1) Replacing the stride-2 downsampling operation in the YOLOv5 backbone with space-to-depth downsampling operations to enhance fine feature extraction for detecting small nodules more comprehensively. (2) Employing an asymptotic feature pyramid network in the YOLOv5 neck to establish connections among feature maps from different paths,thereby enhancing interaction among different hierarchical levels. (3) Introducing global context-aware attention to the end of YOLOv5 neck network for improving the model's ability to understand key features and semantic information of lung nodules from a global perspective. (4) Utilizing the loss rank mining approach to strategically train on hard samples,thereby strengthening the model's discrimination ability. The improved algorithm achieves 96.0％ precision,95.0％ recall rate and 97.3％ average precision on the LUNA16 dataset,which are 14.0％,10.2％ and 12.1％ higher than the original YOLOv5 model,demonstrating its effectiveness for lung nodule detection.

Objective To propose a novel method for verifying the isocenter in radiotherapy based on markers and conduct a preliminary test. Methods A feasibility experiment was conducted on wooden box phantom for radiotherapy resetting. Fifteen groups of displacement data were randomly generated,corresponding to the position deviations of the isocenter in the radiotherapy plan relative to the original isocenter. According to each set of displacement data,with the aid of movable lasers and a CT scanning couch,CT scanning was performed with two sets of markers (3 per set) affixed to the phantom which were corresponding to the original and treatment isocenters,respectively. In the Eclipse treatment planning system,the coordinate data of the original and treatment isocenters were manually verified,and the difference of coordinate data was calculated to obtain the actual displacement value. The treatment isocenter position was finally confirmed by comparing with the actual displacement. In addition,the study attempts to use threshold segmentation algorithm to automatically detect metal markers and obtain coordinate values of the original isocenter on patient-positioned CT images. In the wooden box experiment,the absolute value of the difference between the actual displacement value and the planned displacement value (?d) was used to represent the position accuracy of treatment isocenter,and the deviation value obtained with threshold segmentation algorithm for isocenter detection was ?s. Results The ?d in the X (left-right),Y (superior-inferior) and Z (anterior-superior) directions was (0.83±0.37) mm,0 (0,0.5) mm and (0.45±0.29) mm,respectively. The ?s in the X,Y and Z directions was (0.46±0.22) mm,0 (0,0.5) mm and (0.33±0.29) mm,respectively. The mean values of ?s in 3 directions were all lower than 2 mm,within the range of permissible clinical positioning error. Conclusion The method of automatically verifying treatment isocenter position based on markers is feasible,and the study provides a useful reference for radiotherapy resetting using CT simulator.