Clinical diagnosis for complex disease conditions is a complicated decision process involving systematic inference and differentiation. Artificial Intelligence (AI) models have been a widely established approach to help improve the efficiency of various kinds of clinical decision tasks (e.g., diagnosis, treatment, and prognosis). However, due to the critical requirement of time efficiency, lack of sufficient information, and high probability of comorbid diseases in Outpatient and Emergency Settings (OESs), it is still challenging to build clinically feasible AI models using the free text clinical records in OES for complex disease conditions, such as neurosurgery. Here we propose an AI diagnosis model, named LLM4DEU, for neurosurgery disease differentiations by fine-tuning a large language model (i.e., ChatGLM) using the Department of Neurosurgery, the Beijing Tiantan Hospital OES electronic health records. LLM4DEU obtained state-of-the-art performance on clinical diagnosis with a F1 score of 78.53%, which is superior to five well-known baselines (including deep learning models). In addition, we evaluated the actual performance of the model by case studies on the diagnosis of specific neurosurgical diseases (e.g., subdural hematoma, cerebral hemorrhage, and cerebral infarction). The experimental results show that the LLM4DEU model has significant advantages in diagnosing low-incidence disease conditions, and comparative analyses with clinical experts confirm the predictive power of the model in neurosurgical diagnosis.
- Article type
- Year
- Co-author
In this paper, we introduce a long-term follow-up specific individual searching (SIS) model. This model introduces the concept of node search contributions by considering the characteristics of the network structure. A node search contribution indicates the ability of a certain node to correctly guide the search path and successfully complete an SIS. The influencing factors of node search contributions have three components: the individual influence index, attribute similarity, and node search willingness. On the basis of node search contributions and the PeopleRank idea, this paper proposes an SIS model based on node search contribution values and conducts comparison experiments with several mainstream SIS algorithms in three aspects: the search failure rate, the minimum number of search hops, and the search size. The experimental results verify the advanced nature and operability of the model proposed in this paper, which presents theoretical and practical significance to the quantitative study of the SIS process.
Nowadays, with improvements in the quality of life, people are paying more attention to their health. Traditional Chinese medicine offers great advantages for daily care. In this paper, we present the development of a remote health care system, namely, Chinese Pulse Condition Acquisition System (CPCAS), based on the principle of Chinese pulse diagnosis in Chinese medicine and a wireless sensor network. We designed a remote health care terminal with a mini-pulse collection bench to overcome the challenge of differences in pulse characters of different people. An effective measured pressure control algorithm is proposed to achieve a balance between control accuracy and control time. The special signal conditioning circuit showed good performance in analog pulse signal processing. We also performed significant research to address the challenges of symptom recognition. Other distinctive features of this system include the following: intelligent sensing, a wireless health care network, effective energy management, small size, lightweight, and the ability to be networked for remote management. In this paper, we have introduced the design and implementation of CPCAS. We also demonstrate the use of the system and give evaluations on this system by several experiments. Our results indicate that CPCAS has significant practical feasibility.
In recent years, Compressed Sensing (CS) has been a hot research topic. It has a wide range of applications, such as image processing and speech signal processing owing to its characteristic of removing redundant information by reducing the sampling rate. The disadvantage of CS is that the number of iterations in a greedy algorithm such as Orthogonal Matching Pursuit (OMP) is fixed, thus limiting reconstruction precision. Therefore, in this study, we present a novel Reducing Iteration Orthogonal Matching Pursuit (RIOMP) algorithm that calculates the correlation of the residual value and measurement matrix to reduce the number of iterations. The conditions for successful signal reconstruction are derived on the basis of detailed mathematical analyses. When compared with the OMP algorithm, the RIOMP algorithm has a smaller reconstruction error. Moreover, the proposed algorithm can accurately reconstruct signals in a shorter running time.