Digital Pathology and Artificial Intelligence in Resource-limited Laboratories: A New Frontier for Cancer Diagnosis
Lydia Amarachi Onwuemelem
Department of Medical Laboratory Science, University of Benin, Nigeria.
Olufemi Adesola Adedayo
Department of Mathematics and Statistics, College of Natural Sciences, University of Massachusetts, Amherst, United States
N. Ohale Sandra
*
Department of Pharmacology and Therapeutics, Faculty of Basic Clinical Sciences, College of Medicine, University of Nigeria Nsukka, Nigeria.
Gbadebo Moses Adetoyi
Department of Chemical Pathology and Immunology, University of Ilorin Teaching Hospital, Ilorin, Kwara State, Nigeria.
Morolake Martina Olabanji
Department of Chemical Pathology and Immunology, University of Ilorin Teaching Hospital, Ilorin, Kwara State, Nigeria.
Ndidi Atasie Eboh
Department of Nursing, Stark State College North Canton, Ohio, United States.
*Author to whom correspondence should be addressed.
Abstract
Cancer remains a major global health challenge, with a disproportionate burden in low and middle-income countries (LMICs), with lack of diagnostic services results in late diagnosis and poor outcomes. While histopathology is the standard for cancer diagnosis, its impact in low-resource settings is limited by lack of trained staff, limited resources, and long diagnostic times. These barriers suggest a need for better diagnostic solutions to increase access to timely and accurate cancer diagnosis in LMICs. In this review, we consider the use of digital pathology specifically whole-slide imaging and artificial intelligence (AI) technologies in cancer diagnosis. Results indicate that digital pathology increases access to expert care, facilitates telepathology and optimises diagnostic workflow. AI also assists in automatic tumour detection, tumour grading (such as Ki-67 indication), prediction of molecular biomarkers and prognosis. But reported accuracies are typically on curated data in controlled studies, restricting their broader clinical applicability. The challenges include limited data, algorithmic bias, infrastructural and regulatory hurdles. However, recent developments in federated learning, cost-effective computational models and telepathology integration can provide scalable solutions. But robust validation, workflow integration and continuing investment in digital pathology infrastructure will be needed to achieve this. In general, AI-enabled digital pathology may enhance diagnostic accuracy, efficiency, and global access to cancer care, especially in LMICs, if tailored to the right context and considering ethical aspects.
Keywords: Digital pathology, artificial intelligence, whole-slide imaging, cancer diagnosis, low- and middle-income countries, deep learning, telepathology, computational pathology