A Case Study: Disease Code (ICD-10) Classification in Turkish Medical Summary Dataset

Ozsonmez D. B., ACARMAN T.

6th International Conference on Intelligent Sustainable Systems, ICISS 2023, Tirunelveli, Hindistan, 3 - 04 Şubat 2023, cilt.665 LNNS, ss.537-545 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası: 665 LNNS
  • Doi Numarası: 10.1007/978-981-99-1726-6_41
  • Basıldığı Şehir: Tirunelveli
  • Basıldığı Ülke: Hindistan
  • Sayfa Sayıları: ss.537-545
  • Anahtar Kelimeler: Disease classification, Medical summary, Topic detection, Transfer learning, Turkish
  • Galatasaray Üniversitesi Adresli: Evet

Özet

Information Retrieval (IR) holds significant value due to the opportunities it provides such as atomicity, manageability, accuracy, and relevancy. Topic detection, as a sub-branch of Information Retrieval, receives significant attention in many areas such as marketing, finance, and education. Although it may not be as popular as the aforementioned areas, Medicine is nevertheless an area in which topic detection is significant. In this study, it is aimed to detect the disease classes (International Classification of Diseases (ICD)-10 C-Type) on the Turkish medical summaries’ dataset. The text processing, model creation (FastText and BERT), and the performance measurement steps are realized. As the results, 86% overall F1-Score by using optimized hyperparameters with FastText and 92% by using DistilBERTurk are obtained.