Automated prostate cancer diagnosis based on gleason grading using convolutional neural network

Haotian Xie, Yong Zhang, Jun Wang, Jingjing Zhang, Yifan Ma, Zhaogang Yang

Research output: Contribution to journalArticlepeer-review


Prostate cancer (PCa) is the most common deadly cancer in the United States. The Gleason grading system using histological images is the most powerful diagnostic and prognostic predictor of PCa and is essential in treatment planning for patients. The current standard inspection is evaluating Gleason H&E-stained histopathology images by pathologists. However, it is complicated, time-consuming, and subject to observers. Hence, an automatic classification system is necessary to reduce variations and improve clinical outcomes accuracy. Deep learning (DL) based-methods that automatically learn image features and achieve higher generalization ability have attracted significant attention. However, challenges remain especially using DL to train the whole slide image (WSI), a predominant clinical source in the current diagnostic setting, containing billions of pixels, morphological heterogeneity, and artifacts. To make DL become a reality in the clinical practice, the above difficulties need to be addressed. Hence, we proposed a convolutional neural network (CNN)-based automatic classification method for accurate grading of PCa using whole slide histopathology images. In this paper, a data augmentation method named Patch-Based Image Reconstruction (PBIR) was proposed to reduce the high resolution and increase the diversity of WSIs. In addition, a distribution correction (DC) module was developed to enhance the adaption of pretrained model to the target dataset by adjusting the data distribution. Besides, a Quadratic Weighted Mean Square Error (QWMSE) function was presented to reduce the misdiagnosis caused by equal Euclidean distances. These strategies enabled our method to take advantage of the tremendous amount of clinical information in large and small patches in the labeled dataset. We studied their effects on the classification performance empirically. Our experiments indicated the combination of PBIR, DC, and QWMSE function was necessary for achieving superior expert-level performance, leading to the best results (0.8885 quadratic-weighted kappa coefficient). We expect this system to help pathologists reduce the probability of misdiagnosis and support prostate cancer treatment via an automated, reproducible, and accurate method with consistent standards.

Original languageEnglish (US)
JournalUnknown Journal
StatePublished - Nov 29 2020


  • CNN
  • Diagnosis
  • Gleason score
  • Image classification
  • Loss function
  • Prostate cancer
  • Segmentation

ASJC Scopus subject areas

  • General

Fingerprint Dive into the research topics of 'Automated prostate cancer diagnosis based on gleason grading using convolutional neural network'. Together they form a unique fingerprint.

Cite this