Automatic Prostate Cancer Grading Using Deep Architectures

Prostate cancer is the second most aggressive type of cancer among men aged over 45, and it has a major effect on people's lives. Early diagnosis and grading of prostate cancer from tissue images is necessary. Large scale inter observer reproducibility exists in grading the prostate biopsies. This leads us to move towards a computer based model that can accurately detect and grade the cancerous prostate from non-cancerous one. The paper is focused on deep learning based models to automatically grade the prostate cancer from tissue microarray images. Deep learning models directly learn the features via convolutional layers. Two datasets have been used for implementation of our proposed model, Harvard dataset and Gleason Challenge 2019. Our proposed UNET based architecture is used for training as well as validation and testing. We used four different deep learning models, VGG19, ResNet50, Mobilenetv2 and ResNext50 for our UNET based encoder. With our proposed framework, we have achieved 0.728 and 0.732 average Cohen's kappa with F1 on both datasets respectively. The results show that our proposed UNET based deep learning model shows better performance as compared to other state of the art models.

Automated summary

Prostate cancer is the second most aggressive cancer in men over 45, with a major impact on their lives. The study focuses on deep learning models for automatically grading prostate cancer from tissue images. Utilizing UNet based architecture and four different deep learning models, the proposed framework achieved better performance compared to state of the art models.