ساخت اکانت کاربری رایگان

مقالات پذیرفته شده کنگره

  • صفحه نخست
  • مقالات پذیرفته شده اولین کنگره بین المللی ژنتیک سرطان
بازنشر

  • Artificial Intelligence Strategies in Clinical Pathology and Radiology for Cancer Diagnosis: A Systematic Review

  • Roshanak Roshangah,1,* Ronak Roshangah,2
    1. Faculty of Veterinary Medicine, Science and Research Branch, Islamic Azad University, Tehran, Iran
    2. Faculty of Veterinary Medicine, Sanandaj Branch, Islamic Azad University, Sanandaj, Kurdistan Province, Iran


  • Introduction: Early and accurate cancer diagnosis plays a pivotal role in the prognosis and quality of life of patients. In recent years, artificial intelligence (AI), especially the deep learning branch, has created a significant transformation in diagnostic medicine. Clinical pathology and radiology are two strategic areas that have been receptive to AI technology due to the large volume of image data and the need to recognize complex patterns. The aim of this systematic review is to identify and analyze the main strategies for applying AI in cancer diagnosis in these two disciplines.
  • Methods: This study was conducted in accordance with the PRISMA guidelines. A systematic search was conducted in PubMed, Scopus, Web of Science, and IEEE Xplore databases for English articles published between January 2019 and December 2025. Keywords included a combination of "Artificial Intelligence", "Deep Learning", "Clinical Pathology", "Radiology", and "Cancer Diagnosis". Inclusion criteria was original studies with cohort, case-control or trial design that evaluated at least one AI strategy in cancer diagnosis (breast, lung, prostate, colorectal) in pathology or radiology. Exclusion criteria was case reports, narrative reviews, and studies without extractable data. Finally, 27 articles were included in the analysis.
  • Results: The results from the analysis of 27 included articles showed that AI strategies in the two fields of radiology and clinical pathology are mainly based on convolutional neural networks (CNN) and transfer learning, and cover three main categories of approaches including lesion detection and segmentation, deep feature extraction (radiomics in radiology and morphological analysis in pathology), and multimodal data fusion (multimodal AI). In radiology, deep learning models were able to provide a sensitivity of 88–96% and a false positive reduction of up to 40% compared to radiologists alone. Also, in mammography for breast cancer, algorithms based on the ResNet-50 architecture achieved accuracy equal to or better than two experienced radiologists (average AUC 0.94). The radiomics strategy, reported in several studies, was able to establish a significant relationship between features extracted from PET/CT images and genetic mutation profiles of tumors (such as EGFR in lung cancer and BRAF in melanoma) and showed an AUC above 0.85 in predicting response to immunotherapy. In the field of clinical pathology, whole slide image analysis (WSI) using deep convolutional networks succeeded in identifying microscopic breast cancer metastases in lymph nodes with high accuracy, which in a large study with 2500 slides, its performance was better than that of a general pathologist and equal to that of a specialist pathologist. Also, Gleason-based prostate tumor grading with the help of artificial intelligence models achieved over 90% agreement with the senior pathologist and reduced the interpretation time from an average of 15 minutes to less than 3 minutes. Another emerging strategy, discussed in a few articles, was the prediction of molecular markers directly from hematoxylin-eosin (H&E)-stained images.
  • Conclusion: AI has been able to increase the accuracy of cancer diagnosis and reduce interpretation time with a variety of strategies, including automated lesion detection in radiology, quantitative tissue analysis in pathology, and multimodal data fusion. However, the lack of large standardized databases and the lack of extensive clinical validation are major barriers to implementation. Future studies should focus on prospective multicenter studies and integration of AI models into clinical workflows.
  • Keywords: Artificial intelligence, Clinical pathology, Radiology, Cancer diagnosis, Deep learning

به خانواده بزرگ کنسر ژنتیکس و ژنومیکس سرطان بپیوندید!

ایجاد حساب کاربری رایگان