Mission

Toward building intelligent and authoritative medical brain for empowering real world applications and third party research partners through medical big data processing and machine learning.

  • Raw Data

    Medical Nature Language Processing

  • Computable Data

    Mmedical Knowledge and Data Mining

  • Knowledge Graph

    Knowledge Inference and Machine Learning

  • Application

    Open Research Platform and Ecosystem

Research Directions

  • Medical Natural
    Language Processing

    1. Cross lingual basic natural language processing
    2. Clinical text semantic understanding
    3. Medical knowledge graph construction
  • Medical Image Processing

    1. Fundamental image recognition and segmentation
    2. Studies of medical images driven by applied scenario
    3. Application driven medical image processing model construction
  • Medical Big Data
    Mining and Applications

    1. Big data mining and statistical analysis
    2. Application driven machine learning models
  • Medical Natural Language Processing

    • Chinese Word Segmentation

      A process in which medical texts are rearranged into a word sequence according to certain standards, which is the foundation for further text analysis.

    • Entity Recognition

      Identify medical terms with particular meanings in medical texts, such as “diagnosis names”, “drug names” and “operation names”, and some attribute information having clear semantic meanings such as “time” and “degree”, and build an important preconditions for semantic analysis of medical texts.

    • Entity Relationship Classification

      Identify the relationship between entities in texts, for example, in an chemotherapy instance, identify such items like “time”, “curative effect” of the “chemotherapy plan”, which build an important basis for extraction of structured information.

    • Text Structuralization

      Design models as required by a task, and extract structured task information from the results of entity identification and entity relationship identification.

    • Entity Normalization

      Imitating human normalization, use entity linking technology based on knowledge atlas to normalize clinical data to standard names, for example, to normalize a surgery operation to ICD9, and a diagnosis name to ICD10, thus facilitating analysis of future tasks such as screening and statistics.

    • Knowledge Graph

      Integrate data from multiple Construction sources such as clinical data, medical literature guide, CFDA data and Internet data to establish a large medical knowledge base supporting such tasks of medical statistics, search, analysis and reasoning.

  • Medical Image Processing

    • Clinical Decision Support

      Automatically identify textural anomaly and isolate nodules based on AI technologies, screen cancer at early stage to help doctors with clinical decision-making. Changes in nodules as time goes can be captured to ensure the efficiency of interdisciplinary cooperation.

    • Decision Support In Diagnosis

      Identifying disease areas with the help of images recognition to assist doctors of non-specialty in making basic clinical decisions, currently with an accuracy of 97.5%.

    • Decision Support in Surgery Operations

      Analyzing real time medical images during surgery to assist doctors with hints and alerts obtained by identifying focal forms.

  • Medical Big Data Mining and Applications

    • Descriptive Analysis

      Using real world data statistics to generate data distribution and geo-graphics such as population, regions, diseases, drugs and intervention measures based on mass real clinical data, and support epidemiological studies and analysis of market insight.

    • Correlation Analysis

      Mining, analyzing and validating correlations among different medical factors in real and complex clinical environments to support clinical diagnosis, clinical researches and drug R&D.

    • Inter-Group Comparative Analysis

      Mining and analyzing side effects/adverse reactions of medicine between different patient groups in real clinical processes. Evaluating effectiveness and safety of drugs and treatment plans to improving the quality of diagnosis and treatment as well as expanding potential drug indications and its population.

    • Clinical Treatment Path Mining

      Powered by clinical guidelines, mining treatment patterns and clinical treatment pathways from real world data to support standardization of diagnosis and treatment as well as assisting clinical guideline updating.

    • Data Modeling and Machine Learning

      Through classical machine learning and deep learning technologies, building the scenario driven prediction models such as survival rate prediction of specific diseases, medicine side effect prediction etc.

    • Clinical Decision Support System

      Based on integration of extracted medical knowledge graph, clinical data and digitalized human expert knowledge, building clinical decision support system using machine learning technologies to assist doctors in decision making during pre-diagnosis, mid-diagnosis and post-diagnosis, which can improve both their effectiveness and efficiency.

    • Automatic Knowledge Discovery and Recommendation

      Generating scientific hypothesis through automatic knowledge discovery from real world data with pruning algorithms to reduce the search space, and then doing validations using the data as real world evidences for manual research topic recommendation.