Convolutional neural networks combined with classification algorithms for the diagnosis of periodontitis.

OBJECTIVES: We aim to develop a deep learning model based on a convolutional neural network (CNN) combined with a classification algorithm (CA) to assist dentists in quickly and accurately diagnosing the stage of periodontitis. MATERIALS AND METHODS: Periapical radiographs (PERs) and clinical data were collected. The CNNs including Alexnet, VGG16, and ResNet18 were trained on PER to establish the PER-CNN models for no periodontal bone loss (PBL) and PBL. The CAs including random forest (RF), support vector machine (SVM), naive Bayes (NB), logistic regression (LR), and k-nearest neighbor (KNN) were added to the PER-CNN model for control, stage I, stage II and stage III/IV periodontitis. Heat map was produced using a gradient-weighted class activation mapping method to visualize the regions of interest of the PER-Alexnet model. Clustering analysis was performed based on the ten PER-CNN scores and the clinical characteristics. RESULTS: The accuracy of the PER-Alexnet and PER-VGG16 models with the higher performance was 0.872 and 0.853, respectively. The accuracy of the PER-Alexnet + RF model with the highest performance for control, stage I, stage II and stage III/IV was 0.968, 0.960, 0.835 and 0.842, respectively. Heat map showed that the regions of interest predicted by the model were periodontitis bone lesions. We found that age and smoking were significantly related to periodontitis based on the PER-Alexnet scores. CONCLUSION: The PER-Alexnet + RF model has reached high performance for whole-case periodontal diagnosis. The CNN models combined with CA can assist dentists in quickly and accurately diagnosing the stage of periodontitis.

Deep Reinforcement Learning-Based Power Allocation for Minimizing Age of Information and Energy Consumption in Multi-Input Multi-Output and Non-Orthogonal Multiple Access Internet of Things Systems.

Multi-input multi-output and non-orthogonal multiple access (MIMO-NOMA) Internet-of-Things (IoT) systems can improve channel capacity and spectrum efficiency distinctly to support real-time applications. Age of information (AoI) plays a crucial role in real-time applications as it determines the timeliness of the extracted information. In MIMO-NOMA IoT systems, the base station (BS) determines the sample collection commands and allocates the transmit power for each IoT device. Each device determines whether to sample data according to the sample collection commands and adopts the allocated power to transmit the sampled data to the BS over the MIMO-NOMA channel. Afterwards, the BS employs the successive interference cancellation (SIC) technique to decode the signal of the data transmitted by each device. The sample collection commands and power allocation may affect the AoI and energy consumption of the system. Optimizing the sample collection commands and power allocation is essential for minimizing both AoI and energy consumption in MIMO-NOMA IoT systems. In this paper, we propose the optimal power allocation to achieve it based on deep reinforcement learning (DRL). Simulations have demonstrated that the optimal power allocation effectively achieves lower AoI and energy consumption compared to other algorithms. Overall, the reward is reduced by 6.44% and 11.78% compared the to GA algorithm and random algorithm, respectively.

Application of antimicrobial photodynamic therapy to treat subgingival multidrug-resistant bacterial infections in ICU patients.

BACKGROUND: Drug-resistant bacterial infections have received much attention in recent years. Antimicrobial photodynamic therapy (aPDT) is an effective antimicrobial strategy. This study aimed to evaluate the therapeutic effect of methylene blue (MB)-mediated aPDT against subgingival multidrug-resistant (MDR) bacterial infections in intensive care unit (ICU) patients. METHODS: Eighty-three patients who were hospitalized in the ICU of the Second Affiliated Hospital of Nanchang University from July 2019 to June 2021 were selected. The intraoral partitioned control test was conducted. Teeth that met the criteria were selected from different quadrants of the same patient, randomly divided into three groups, namely, A, B, and C, and treated with aPDT, chlorhexidine gargle, or normal saline. The counts of MDR bacteria in the gingival crevicular fluid were assessed in the different groups at different time points before and after treatment. RESULTS: The MDR bacterial count decreased immediately after aPDT and was significantly different from that in the chlorhexidine gargle rinse group and the normal saline rinse group (P<0.05). There was no significant difference among the three groups at 6, 12, and 24 hours after treatment (P>0.05). CONCLUSION: aPDT can be used to treat subgingival MDR bacterial infections, but the long-term effects of treatment need to be further studied.

Robust Radiation Sources Localization Based on the Peak Suppressed Particle Filter for Mixed Multi-Modal Environments.

This paper addresses a detection problem where sparse measurements are utilized to estimate the source parameters in a mixed multi-modal radiation field. As the limitation of dimensional scalability and the unimodal characteristic, most existing algorithms fail to detect the multi-point sources gathered in narrow regions, especially with no prior knowledge about intensity and source number. The proposed Peak Suppressed Particle Filter (PSPF) method utilizes a hybrid scheme of multi-layer particle filter, mean-shift clustering technique and peak suppression correction to solve the major challenges faced by current existing algorithms. Firstly, the algorithm realizes sequential estimation of multi-point sources in a cross-mixed radiation field by using particle filtering and suppressing intensity peak value, while existing algorithms could just identify single point or spatially separated point sources. Secondly, the number of radioactive sources could be determined in a non-parametric manner as the fact that invalid particle swarms would disperse automatically. In contrast, existing algorithms either require prior information or rely on expensive statistic estimation and comparison. Additionally, to improve the prediction stability and convergent performance, distance correction module and configuration maintenance machine are developed to sustain the multimodal prediction stability. Finally, simulations and physical experiments are carried out in aspects such as different noise level, non-parametric property, processing time and large-scale estimation, to validate the effectiveness and robustness of the PSPF algorithm.