Automated diagnosis of prostate cancer in multi-parametric MRI based on multimodal convolutional neural networks.

Automated diagnosis of prostate cancer in multi-parametric MRI based on multimodal convolutional neural networks. Phys Med Biol. 2017 Jun 05;: Authors: Le MH, Chen J, Wang L, Wang Z, Liu W, Cheng KT, Yang X Abstract Automated methods for prostate cancer (PCa) diagnosis in multi-parametric magnetic resonance images (MP-MRI) are critical for increasing the survival rate of patients and alleviating requirements in radiology reading and reducing risk of over-/under-treatment. This paper presents an automated method based on multimodal convolutional neural networks (CNNs) for two PCa diagnosis tasks: 1) distinguishing between cancer and noncancerous tissues, 2) distinguishing between clinically significant (CS) and indolent PCa. Specifically, our multimodal CNNs effectively fuse apparent diffusion coefficients (ADC) and T2 weighted images (T2WI) of MP-MRI. New back-propagate error is designed for our multimodal CNNs to enforce optimized classification accuracy and consistent features for both modalities during CNN feature training. Such enforcement enables better fusion results than existing methods as the feature learning process of both modalities are mutually guided by each other, jointly facilitating CNN to 'see' true visual patterns of PCa. The classification results of multimodal CNNs are further combined with the results based on handcrafted features using a SVM classifier. To achieve a satisfactory accuracy for clinical usage, we ...
Source: Physics in Medicine and Biology - Category: Physics Authors: Tags: Phys Med Biol Source Type: research