本研究では，ソーシャルメディア上のがん患者の記述や語りを対象として，自然言語処理技術を用いて疾患や治療に関連する自覚症状や悩みのエピソードを抽出する手法を開発した．具体的には，"身体の苦痛"や"就労・経済的負担"など複合的に発生する悩みを捉える高性能なマルチラベル分類器，実践知の把握や感情分析を行う手法，手足症候群シグナルや日常生活に支障を及ぼす有害事象シグナルの抽出器の構築に成功した．さらに，患者症状訴えを含む異なるテキストソース（薬局薬歴等）に対してもこれらの手法を活用し，その臨床的有用性を示した．
In this study, episodes of subjective symptoms and concerns related to diseases and treatments were effectively extracted from narratives of cancer patients on social media, using natural language processing (NLP) technology. The research specifically targeted unique concerns of cancer patients, such as "physical concerns" and "work/financial burdens," which led to the development of a high-performance multi-label classifier. Moreover, an innovative extractor was designed to identify signals for hand-foot syndrome and adverse events with a significant impact on daily-life activities. The clinical utility of this extractor was demonstrated through its application to various text sources, including pharmaceutical care records containing detailed patients’ concerns. The significance of this research lies in its development of a method to extract data from non-medical texts using cutting-edge NLP technology, capturing concerns and practical knowledge from the patients' perspective.

研究種目 : 基盤研究 (B) (一般)
研究期間 : 2021～2023
課題番号 : 21H03170
研究分野 : 健康情報学