AI Bibliography
AI Bibliography |
|
 |
 |
 |
|
Tognola, G., Murri, A., & Cuda, D. 2018, Cognitive computing for the automated extraction and meaningful use of health data in narrative medical notes: An application to the clinical management of hearing impaired aged patients. Paper presented at 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI). |
| Resource type: Proceedings Article BibTeX citation key: Tognola2018 View all bibliographic details  |
Categories: Artificial Intelligence, Cognitive Science, Complexity Science, Computer Science, Data Sciences, Decision Theory, General, Medical science Subcategories: Big data, Decision making, Deep learning, Human decisionmaking, Informatics, Machine learning, Machine recognition, Neural nets, Neurosymbolic, Psychology of human-AI interaction, Systems theory Creators: Cuda, Murri, Tognola Publisher: Collection: 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI) |
| Attachments |
| Abstract |
|
It is currently estimated that 80% of health data, such as that coming from medical transcripts, medical notes, lab results, email messages, attachments, is unstructured. Being unstructured, this information cannot be processed through automated PC procedures but requires the interpretation of human beings. We propose a cognitive computing system to extract automatically meaningful health information from the textual documents of the patient folder and merge this information into a structured data frame. The system was tested on the medical documents generated in an audiological outpatient hospital service; the data corpus consisted of the documents and reports generated longitudinally from the enrollment visit to the last available follow-up of hearing impaired aged patients treated with cochlear implants (CI). The system is based on an Information Extraction (IE) module to extract meaningful health information, a couple of ontologies to interpret the meaning and classify the extracted information into a logical hierarchy and an ad hoc developed structured data frame to gather the information. The system was designed to be compliant with the clinical best practices of the audiological/ENT (Ear-Nose-Throat) medical domains to ensure its ease of use in the real practice. The performance was assessed by measuring the percentage of information correctly extracted by the system against the one manual extracted by two experts. The accuracy of the system was very good (recall= 0.78; precision=3D0.95).
|