2019 Fiscal Year Annual Research Report
From Reports to Knowledge for Patient Safety Improvement through Advancements in Artificial Intelligence
| Project/Area Number |
18H03336
|
|---|
| Research Institution | St. Luke's International University |
|---|
Principal Investigator |
ウォン スイー 聖路加国際大学, 専門職大学院公衆衛生学研究科(公衆衛生大学院), 准教授 (70791599)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
林 邦好 聖路加国際大学, 専門職大学院公衆衛生学研究科(公衆衛生大学院), 講師 (00793217)
笹野 遼平 名古屋大学, 情報学研究科, 准教授 (70603918)
|
|---|
| Project Period (FY) |
2018-04-01 – 2023-03-31
|
| Keywords | WHO Patient Safety / Artificial Intelligence / Incident reports / Deep learning (DL) / Named entity recognition / Adverse drug events (ADEs) |
|---|
| Outline of Annual Research Achievements |
We have successfully developed annotation guideline, created annotated data, developed NER models, and explored the next generation patient safety system design. We have also published our ongoing findings to various reputable international conferences and journals. Our AI patient safety research also gained local and international attentions. For instance, Dr Wong provided an invited talk on AI for patient safety at the 7th Health AI Acceleration Consortium Meeting at the Ministry of Health, Labour and Welfare. Dr Shin Ushiro communicated our ongoing project outcomes as an example project of the WHO third patient safety challenges (medication without harm) in the WHO Salzburg Seminar.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
In the last fiscal year, the research team focused on the validation of medication error annotation method (Task 1) and the development named entity recognition (NER) model framework for incident reports learning (Task 2). Using our St. Luke’s and JQ incident report data, our team validated the proposed annotation guideline for incident reports and created annotated incident reports data for NER model training and validation. We evaluated multiple NER program settings that are suitable for incident report learning. We have come up with a draft design of the prototype NER system (Task 3) by incorporating the NER model prediction as interactive outputs to aid incident reporting. These activities fulfilled the original project progress and plans.
|
| Strategy for Future Research Activity |
Using our validated annotation method, we have created hundreds of annotated data using incident report from JQ. Led by Dr Sasano, our team will refine the appropriate NER models and evaluate the model performance. In the next fiscal year, we will develop the prototype NER system for patient safety and evaluate the performance of the system. To communicate our outcome with a wider community, we will continue the current efforts in connecting with the local and global patient safety and AI communities (including WHO patient safety). We will also write up research reports as conference/journal papers to document the key research outcomes. We hope our research would make significant impact to the local and global patient safety practice.
|
Research Products
(6 results)
