한국방사선학회논문지 (Journal of the Korean Society of Radiology)

  • 제18권5호
  • /
  • Pages.551-556
  • /
  • 2024
  • /
  • 1976-0620(pISSN)
  • /
  • 2384-0633(eISSN)
한국방사선학회 (The Korean Society of Radiology)
권호 표지
DOI QR코드

DOI QR Code

딥러닝을 기반으로 한 CAD 시스템의 갑상샘 질환의 진단 유용성

Evaluation of Diagnostic Usefulness of Thyroid Lesions of Deep Learning-based CAD System

  • 강채원 (동의대학교 대학원 보건의과학과) ;
  • 이효영 (동의대학교 대학원 보건의과학과)
  • Chae Won Kang (Department of Biomedical Health Science, Graduate School of Dong-Eui University) ;
  • Hyo Yeong Lee (Department of Biomedical Health Science, Graduate School of Dong-Eui University)
  • 투고 : 2024.08.23
  • 심사 : 2024.10.31
  • 발행 : 2024.10.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

본 연구는 인공지능 기반 컴퓨터 진단 보조 시스템(CAD)인 S-DetectTM를 통해 진단된 갑상샘 병변과 세침흡인 검사(FNAB) 결과를 비교하여 분석하고, 진단의 일치도와 정확도를 평가하고자 한다. 2023년 5월부터 2023년 9월까지 경남 소재 N 병원 내과에서 60명의 환자를 대상으로 후향적 연구를 수행하였다. S-DetectTM를 사용하여 갑상샘 결절의 초음파 소견과 악성 위험도를 분석하고, 이를 세침흡인 검사 결과와 비교하여 정확도를 확인하였다. S-Detect 방법과 세침흡인 검사 방법 간의 민감도, 특이도, 정확도, 양성예측도 및 음성 예측도를 분석하였으며, 두 방법 간의 진단 일치도를 Kappa 분석을 통해 확인하였다. S-Detect 분석결과는 민감도 90.5%, 특이도 83.2%, 정확도 88.3%, 양성 예측도 80.7%, 음성 예측도 92.7%로 나타났다. 또한, S-Detect 방법과 세침흡인검사 방법 간의 진단 일치도 분석 결과, Kappa 값이 0.719(p<0.05)로 높게 나타났으며, 두 방법 간에 유사한 일치도를 보였다. 따라서, 인공지능 기반 컴퓨터 진단 보조 시스템(CAD)인 S-Detect는 갑상샘 병변에서 악성 결절과 양성 결절을 구별하는데 유용하며, 갑상선 세침흡인 검사 전에 적절히 활용하면 불필요한 조직 검사를 줄일 수 있을 것으로 생각한다.

This study aims to evaluate the diagnostic concordance and accuracy by comparing thyroid lesions diagnosed with the artificial intelligence-based computer-aided diagnosis (CAD) system, S-DetectTM, to the results of fine-needle aspiration biopsy(FNAB). A retrospective study was conducted involving 60 patients at N Hospital in Gyeongnam from May 2023 to September 2023. The study used S-DetectTM to analyze ultrasound findings and malignancy risk of thyroid nodules and compared these findings with FNAB results to determine accuracy. The study assessed the sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) of S-DetectTM and evaluated the diagnostic concordance between the two methods using Kappa analysis. S-DetectTM demonstrated a sensitivity of 90.5%, specificity of 83.2%, accuracy of 88.3%, PPV of 80.7%, and NPV of 92.7%. The Kappa value for diagnostic agreement between S-DetectTM and FN AB was 0.719 (p<0.05), indicating a high level of agreement between the methods. Therefore, the CAD system S-DetectTM proves valuable in distinguishing between malignant and benign thyroid lesions and could reduce unnecessary tissue examinations when used appropriately before thyroid fine-needle aspiration.

키워드

과제정보

이 논문은 2024년 동의대학교 교내 연구비에 의해 연구되었음.(202401240001)

참고문헌

  1. J. H. Moon, M. K. Hyun, J. Y. Lee, et al., "Prevalence of thyroid nodules and their associated clinical parameters : a large-scale, multicenter-based health checkup study", Korean journal of internal medicine, Vol. 33, No. 4, pp. 753-762, 2018. http://dx.doi.org/10.3904/kjim.2015.273
  2. National Cancer Information Center, From URL; https://www.cancer.go.kr/lay1/S1T639C641/contents.do
  3. F. Bray, J. Ferlay, I. Soerjomataram, et al., "Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries", CA: A Cancer Journal for Clinicians, Vol. 68, No. 6 pp. 394-424, 2018. http://dx.doi.org/10.3322/caac.21492
  4. B. R. Haugen, "2015 American Thyroid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer: What is new and what has changed?", Cancer, Vol. 123, No. 3, pp. 372-381, 2017. http://dx.doi.org/10.1002/cncr.30360
  5. M. Dighe, R. Barr, J. Bojunga, et al., "Thyroid ultrasound: State of the art. part 2 - focal thyroid lesions", Medical Ultrasonography, Vol. 19, No. 2, pp. 195-210, 2017. http://dx.doi.org/10.11152/mu-999
  6. J. K. Hoang, W. D. Middleton, A. E. Farjat, et al., "Interobserver Variability of Sonographic Features Used in the American College of Radiology Thyroid Imaging Reporting and Data System", American Journal of Roentgenology, Vol. 211, No. 1, pp. 162-167, 2018. http://dx.doi.org/10.2214/AJR.17.19192
  7. H. J. Lee, D. Y. Yoon, Y. L. Seo, et al., "Intraobserver and Interobserver Variability in Ultrasound Measurements of Thyroid Nodules", Journal of Ultrasound Medicine, Vol. 37, No. 1, pp. 173-178, 2018. http://dx.doi.org/10.1002/jum.14316
  8. Y J Kim, Y S Choi, et al., "Deep convolutional neural network for classification of thyroid nodules on ultrasound: Comparison of the diagnostic performance with that of radiologists" European Journal of Radiology ,Vol. 152, pp.110335, 2022. https://doi.org/10.1016/j.ejrad.2022.110335
  9. J. H. Lee, Y. K. Seong, C. H. Chang, et al., "Computer- aided lesion diagnosis in B-mode ultrasound by border irregularity and multiple sonographic features", Computer Science Medicine, Vol. 8670, 2013. https://doi.org/10.1117/12.2007452
  10. S. Z. Ali, Z. W. Baloch, B. Cochand-Priollet, et al., "The 2023 Bethesda System for Reporting Thyroid Cytopathology", Thyroid, Vol. 12, No. 5, 2023. https://doi.org/10.1089/thy.2023.0141
  11. Y. Chang, A. K. Paul, N. Kim, et al., "Computer-aided diagnosis for classifying benign versus malignant thyroid nodules based on ultrasound images: A comparison with radiologist-based assessments", Medical Physics, Vol. 43, No. 1 pp. 554-567, 2016. http://dx.doi.org/10.1118/1.4939060
  12. E. J. Ha, D. G. Na, Y. H. Lee, et al., "A Multicenter Prospective Validation Study for the Korean Thyroid Imaging Reporting and Data System in Patients with Thyroid Nodules", Korean Journal Radiology, Vol. 17, No. 5, pp. 811-821,
  13. J. H. Shin, J. H. Baek, J. Chung, et al., "Ultrasonography Diagnosis and Imaging-Based Management of Thyroid Nodules: Revised Korean Society of Thyroid Radiology Consensus Statement and Recommendations", Korean Journal Radiology, Vol. 17 No. 3, pp. 370-395, 2016. http://dx.doi.org/10.3348/kjr.2016.17.3.370
  14. T. Hirning, I. Zuna, D. Schlaps, et al., "Quantification and classification of echographic findings in the thyroid gland by computerized B-mode texture analysis", European Journal Radiology, Vol. 9, No. 4, pp. 244-247, 1989.
  15. C. L. Chng, H. C. Tan, C. W. Too, et al., "Diagnostic performance of ATA, BTA and TIRADS sonographic patterns in the prediction of malignancy in histologically proven thyroid nodules", Singapore Medical Journal, Vol. 59, No. 11, pp. 578-583, 2018. http://dx.doi.org/10.11622/smedj.2018062
  16. U. R. Acharya, O. Faust, S. V. Sree, et al., "Cost-effective and non-invasive automated benign and malignant thyroid lesion classification in 3D contrast-enhanced ultrasound using combination of wavelets and textures: a class of ThyroScan™ algorithms", Technology in Cancer Research & Treatment. Vol. 10, No. 4, pp. 371-380, 2011. http://dx.doi.org/10.7785/tcrt.2012.500214
  17. E. Szczepanek-Parulska, K. Wolinski, K. Dobruch-Sobczak, et al., "S-Detect Software vs. EU-TIRADS Classification: A Dual-Center Validation of Diagnostic Performance in Differentiation of Thyroid Nodules", Journal of Clinical Medicine, Vol. 9, No. 8, pp. 2495-2576, 2020. http://dx.doi.org/10.3390/jcm9082495
  18. S. Martina, C. Luca, C. Arturo, et al., "Texture analysis and machine learning to characterize suspected thyroid nodules and differentiated thyroid cancer: Where do we stand?", European Journal of Radiology, Vol. 99, pp.1-8, 2018. http://dx.doi.org/10.1016/j.ejrad.2017.12.004