Particle filter-based parameter estimation algorithm for prognostic risk assessment of progression in non-small cell lung cancer.

Non-small cell lung cancer (NSCLC) is a malignant tumor that threatens human life and health. The development of a new NSCLC risk assessment model based on electronic medical records has great potential for reducing the risk of cancer recurrence. In this process, machine learning is a powerful method for automatically extracting risk factors and indicating impact weights for NSCLC deaths. However, when the number of samples reaches a certain value, it is difficult for machine learning to improve the prediction accuracy, and it is also challenging to use the characteristic data of subsequent patients effectively. Therefore, this study aimed to build a postoperative survival risk assessment model for patients with NSCLC that updates the model parameters and improves model accuracy based on new patient data. The model perspective was a combination of particle filtering and parameter estimation. To demonstrate the feasibility and further evaluate the performance of our approach, we performed an empirical analysis experiment. The study showed that our method achieved an overall accuracy of 92% and a recall of 71% for deceased patients. Compared with traditional machine learning models, the accuracy of the model estimated by particle filter parameters has been improved by 2%, and the recall rate for dead patients has been improved by 11%. Additionally, this study outcome shows that this method can better utilize subsequent patients' characteristic data, be more relevant to different patients, and help achieve precision medicine.

[A Fluorescence Diffusion Optical Tomography System Based on Lattice Boltzmann Forward Model].

Fluorescence Diffuse Optical Tomography (FDOT) is significant for biomedical applications, such as medical diagnostics, drug research. The fluorescence probe distribution in biological tissues can be quantitatively and non-invasively obtained via FDOT, achieving targets positioning and detection. In order to reduce the cost of FDOT, this study designs a FDOT system based on Lattice Boltzmann forward model. The system is used to realize two functions of light propagation simulation and FDOT reconstruction, and is composed of a parameter module, an algorithm module, a result display module and a data interaction module. In order to verify the effectiveness of the platform, this study carries out the light propagation simulation experiment and the FDOT reconstruction experiment, respectively comparing the Monte Carlo (MC) light propagation simulation results and the real position of the light source to be reconstructed. Experiments show that the proposed FDOT system has good reliability and has a high promotion value.

[A Method for Fluorescent Diffuse Optical Tomography Based on Lattice Boltzmann Forward Model on GPU Parallelization].

Fluorescent Diffuse Optical Tomography (FDOT) is an emerging imaging method with great prospects in fields of biology and medicine. However, the current solutions to the forward problem in FDOT are time consuming, which greatly limit the application. We proposed a method for FDOT based on Lattice Boltzmann forward model on GPU to greatly improve the computational efficiency. The Lattice Boltzmann Method (LBM) was used to construct the optical transmission model. This method separated the LBM into collision, streaming and boundary processing processes on GPUs to perform the LBM efficiently, which were local computational and inefficient on CPU. The feasibility of the proposed method was verified by the numerical phantom and the physical phantom experiments. The experimental results showed that the proposed method achieved the best performance of a 118-fold speed up under the precondition of simulation accuracy, comparing to the diffusion equation implemented by Finite Element Method (FEM) on CPU. Thus, the LBM on the GPU may efficiently solve the forward problem in FDOT.

[A Lattice Boltzmann Forward Model for Fast Fluorescence Diffuse Optical Tomography].

Fluorescent Diffuse Optical Tomography (FDOT), as a new imaging technology, can achieve three-dimensional quantitative functional imaging of probe in biological tissues, and has wide application value in biomedicine. Forward model which describes the photon propagation within a biological tissue is a prerequisite for implementing FDOT and determines the performance of FDOT. To further improve the efficiency of FDOT, this paper proposes a new forward model based on the Lattice Boltzmann (LB) method derived from the discretization of radiation transfer equation and applies it to FDOT. The experimental results of numerical simulation and physical phantom show that the LB-based forward model proposed in this paper can increase the imaging speed of FDOT by about 5 times compared with the traditional diffusion equation method, without reducing its imaging quality.

Blockchain-based Electronic Medical Record Security Sharing Scheme.

To solve the difficulty of medical data sharing in traditional medical information systems, we proposed an electronic medical record secure-sharing scheme based on the Blockchain technique. The encrypted text of the patient's electronic medical record is stored in the cloud server while the metadata of the medical record and access strategy is stored in the blockchain system. We employed smart contracts in the blockchain system to achieve user rights management. We used the decentralized, tamper-proof, and traceable features of the blockchain to realize the safe sharing of electronic medical records. The experimental results of security analysis show that the method can defend against potential network attacks while satisfying patient privacy protection and confidentiality. This study verifies the feasibility and great operating efficiency of the blockchain-based electronic medical record security sharing scheme.Clinical relevance- Our proposed blockchain-based electronic medical record-sharing scheme has great potential for the safe access of third-party users to patient data.

Explainable and efficient deep early warning system for cardiac arrest prediction from electronic health records.

Cardiac arrest (CA) is a fatal acute event. The development of new CA early warning system based on time series of vital signs from electronic health records (EHR) has great potential to reduce CA damage. In this process, recursive architecture-based deep learning, as a powerful tool for time series data processing, enables automatically extract features from various monitoring clinical parameters and to further improve the performance for acute critical illness prediction. However, the unexplainable nature and excessive time caused by black box structure with poor parallelism are the limitations of its development, especially in the CA clinical application with strict requirement of emergency treatment and low hidden dangers. In this study, we present an explainable and efficient deep early warning system for CA prediction, which features are captured by an efficient temporal convolutional network (TCN) on EHR clinical parameters sequence and explained by deep Taylor decomposition (DTD) theoretical framework. To demonstrate the feasibility of our method and further evaluate its performance, prediction and explanation experiments were performed. Experimental results show that our method achieves superior CA prediction accuracy compared with standard national early warning score (NEWS), in terms of overall AUROC (0.850 Vs. 0.476) and F1-Score (0.750 Vs. 0.450). Furthermore, our method improves the interpretability and efficiency of deep learning-based CA early warning system. It provides the relevance of prediction results for each clinical parameter and about 1.7 times speed enhancement for system calculation compared with the long short-term memory network.

Contextual embedding bootstrapped neural network for medical information extraction of coronary artery disease records.

Coronary artery disease (CAD) is the major cause of human death worldwide. The development of new CAD early diagnosis methods based on medical big data has a great potential to reduce the risk of CAD death. In this process, neural network (NN), as a powerful tool for electronic medical record (EMR) processing, enables extract structured data accurately to unlock medical information and to further improve CAD diagnosis. However, the excessive time and labor caused by dataset's annotation is the main limitation of its application, especially on the CAD records situation with large natural language text and biomedical professional content. In this study, we present an annotation cost saving NN approach for CAD records, which is bootstrapped by deep language model with contextual embedding pre-trained on large unannotated CAD corpus. To demonstrate the feasibility and to further evaluate the performance of our approach, we performed pre-training experiment and term classification experiment, by using the unannotated and annotated CAD records, respectively. The results showed that our contextual embedding bootstrapped NN for CAD records has better performance under the condition of annotations reduction.