Scholarlink

<p>Timely and accurate diagnosis of common illnesses such as colds, flu, and infectious diseases plays a critical role in improving patient outcomes and preventing mild illnesses from escalating to severe complications. Despite significant medical advances, traditional diagnostic processes still rely heavily on human expertise, which can be time-consuming and prone to human error – especially when symptoms overlap between multiple diseases. In recent years, the rapid evolution of artificial intelligence and deep learning has opened up new horizons in healthcare, enabling the creation of intelligent systems that can help clinicians make faster, more consistent, and data-driven diagnostic decisions. This research focuses on developing a learning model for diagnosing cold, flu, and infectious diseases. Based on the symptoms and tests associated with each of these illnesses, the goal is to create a system that can simulate and predict these diseases with high accuracy. For this purpose, improved Long Short-Term Memory (iLSTM) recurrent neural networks are utilized, leveraging medical data—including features such as temperature, blood pressure, blood sugar, and other symptoms—to train the model. Additionally, test images are processed using the Easy Optical Character Recognition (EasyOCR) method to extract influential features related to disease classification. Once trained, the iLSTM model will be able to analyze patient data, which is structured in an Excel format, and after training, it will be capable of diagnosing diseases based on new inputs, assisting doctors in making fast and accurate diagnoses. To facilitate patient interaction and data retrieval, a Question-Answering (QA) system is designed to ask users for relevant features required for prediction, process test images (if available), and ultimately announce the diagnosis. The QA system is built using Natural Language Processing (NLP) techniques. In this approach, sentences are preprocessed and provided to the model to generate appropriate questions. The results of the experiments and comparison of the iLSTM model with other developed models show the superiority of the presented model. This system can serve as an auxiliary tool to reduce human error and accelerate disease treatment. The code and data are publicly available at <a href="https://github.com/ArmaghanNikfar/Disease-Diagnosis-in-a-Multimodal-System">https://github.com/ArmaghanNikfar/Disease-Diagnosis-in-a-Multimodal-System</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel LSTM-based multimodal system for robust diagnosis of respiratory infectious diseases

  • 期刊论文

  • 发表日期：

    2025年12月4日

查看全文
  • Armaghan Nikfar,
  • Amir Shoja,
  • Sasan Harifi,
  • Javad Mohammadzadeh,
  • Belikova Ekaterina Alexandrovna

摘要

Timely and accurate diagnosis of common illnesses such as colds, flu, and infectious diseases plays a critical role in improving patient outcomes and preventing mild illnesses from escalating to severe complications. Despite significant medical advances, traditional diagnostic processes still rely heavily on human expertise, which can be time-consuming and prone to human error – especially when symptoms overlap between multiple diseases. In recent years, the rapid evolution of artificial intelligence and deep learning has opened up new horizons in healthcare, enabling the creation of intelligent systems that can help clinicians make faster, more consistent, and data-driven diagnostic decisions. This research focuses on developing a learning model for diagnosing cold, flu, and infectious diseases. Based on the symptoms and tests associated with each of these illnesses, the goal is to create a system that can simulate and predict these diseases with high accuracy. For this purpose, improved Long Short-Term Memory (iLSTM) recurrent neural networks are utilized, leveraging medical data—including features such as temperature, blood pressure, blood sugar, and other symptoms—to train the model. Additionally, test images are processed using the Easy Optical Character Recognition (EasyOCR) method to extract influential features related to disease classification. Once trained, the iLSTM model will be able to analyze patient data, which is structured in an Excel format, and after training, it will be capable of diagnosing diseases based on new inputs, assisting doctors in making fast and accurate diagnoses. To facilitate patient interaction and data retrieval, a Question-Answering (QA) system is designed to ask users for relevant features required for prediction, process test images (if available), and ultimately announce the diagnosis. The QA system is built using Natural Language Processing (NLP) techniques. In this approach, sentences are preprocessed and provided to the model to generate appropriate questions. The results of the experiments and comparison of the iLSTM model with other developed models show the superiority of the presented model. This system can serve as an auxiliary tool to reduce human error and accelerate disease treatment. The code and data are publicly available at https://github.com/ArmaghanNikfar/Disease-Diagnosis-in-a-Multimodal-System.