International Journal of General Medicine (Dec 2023)

Use of Novel Open-Source Deep Learning Platform for Quantification of Ki-67 in Neuroendocrine Tumors – Analytical Validation

  • Zehra T,
  • Shams M,
  • Ali R,
  • Jafri A,
  • Khurshid A,
  • Erum H,
  • Naqvi H,
  • Abdul-Ghafar J

Journal volume & issue
Vol. Volume 16
pp. 5665 – 5673

Abstract

Read online

Talat Zehra,1 Mahin Shams,2 Rabia Ali,3 Asad Jafri,3 Amna Khurshid,3 Humaira Erum,3 Hanna Naqvi,3 Jamshid Abdul-Ghafar4 1Pathology Department, Jinnah Sindh Medical University, Karachi, Pakistan; 2Pathology Department, United Medical and Dental College, Karachi, Pakistan; 3Histopathology Department, Liaquat National Hospital, Karachi, Pakistan; 4Department of Pathology and Clinical Laboratory, French Medical Institute for Mothers and Children (FMIC), Kabul, AfghanistanCorrespondence: Jamshid Abdul-Ghafar, Department of Pathology and Clinical Laboratory, French Medical Institute for Mothers and Children (FMIC), Kabul, Afghanistan, Tel+93792827287, Email [email protected]: Neuroendocrine tumors (NETs) represent a diverse group of neoplasms that arise from neuroendocrine cells, with Ki-67 immunostaining serving as a crucial biomarker for assessing tumor proliferation and prognosis. Accurate and reliable quantification of Ki-67 labeling index is essential for effective clinical management.Methods: We aimed to evaluate the performance of open-source/open-access deep learning cloud-native platform, DeepLIIF (https://deepliif.org), for the quantification of Ki-67 expression in gastrointestinal neuroendocrine tumors and compare it with the manual quantification method.Results: Our results demonstrate that the DeepLIIF quantification of Ki-67 in NETs achieves a high degree of accuracy with an intraclass correlation coefficient (ICC) = 0.885 with 95% CI (0.848– 0.916) which indicates good reliability when compared to manual assessments by experienced pathologists. DeepLIIF exhibits excellent intra- and inter-observer agreement and ensures consistency in Ki-67 scoring. Additionally, DeepLIIF significantly reduces analysis time, making it a valuable tool for high-throughput clinical settings.Conclusion: This study showcases the potential of open-source/open-access user-friendly deep learning platforms, such as DeepLIIF, for the quantification of Ki-67 in neuroendocrine tumors. The analytical validation presented here establishes the reliability and robustness of this innovative method, paving the way for its integration into routine clinical practice. Accurate and efficient Ki-67 assessment is paramount for risk stratification and treatment decisions in NETs and AI offers a promising solution for enhancing diagnostic accuracy and patient care in the field of neuroendocrine oncology.Keywords: digital image analysis, histopathology, Ki-67 proliferation index, neuroendocrine tumors, machine learning

Keywords