Evolutionary pruning of transfer learned deep convolutional neural network for breast cancer diagnosis in digital breast tomosynthesis

Abstract

Deep learning models are highly parameterized, resulting in difficulty in inference and transfer learning for image recognition tasks. In this work, we propose a layered pathway evolution method to compress a deep convolutional neural network (DCNN) for classification of masses in digital breast tomosynthesis (DBT). The objective is to prune the number of tunable parameters while preserving the classification accuracy. In the first stage transfer learning, 19 632 augmented regions-of-interest (ROIs) from 2454 mass lesions on mammograms were used to train a pre-trained DCNN on ImageNet. In the second stage transfer learning, the DCNN was used as a feature extractor followed by feature selection and random forest classification. The pathway evolution was performed using genetic algorithm in an iterative approach with tournament selection driven by count-preserving crossover and mutation. The second stage was trained with 9120 DBT ROIs from 228 mass lesions using leave-one-case-out cross-validation. The DCNN was reduced by 87% in the number of neurons, 34% in the number of parameters, and 95% in the number of multiply-and-add operations required in the convolutional layers. The test AUC on 89 mass lesions from 94 independent DBT cases before and after pruning were 0.88 and 0.90, respectively, and the difference was not statistically significant (p > 0.05). The proposed DCNN compression approach can reduce the number of required operations by 95% while maintaining the classification performance. The approach can be extended to other deep neural networks and imaging tasks where transfer learning is appropriate.

Deep Learning and Artificial Intelligence for the Determination of the Cervical Vertebra Maturation Degree from Lateral Radiography

“…The success that brought the breakthrough in the ImageNet competition in 2012 is related to the efficient use of Graphical Processing Units (GPU), data augmentation, rectified linear units, new dropout regularization and deep network architecture, where convolution operations were repeated multiple times between max-pooling operations. Since then, there has been a trend to make this style of network increasingly deeper through the use of small (3 × 3) convolutional filters such as VGG architecture [ 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 ].…”

Artificial intelligence in the interpretation of breast cancer on MRI

“…Deep learning through convolutional neural networks has now been applied to the characterization task in the diagnostic workup of breast tumors in mammography/tomosynthesis, ultrasound, and MRI . In the task of distinguishing between malignant and benign breast lesions, transfer learning has been used with a pretrained CNN either through feature extraction or fine tuning.…”

Deep learning in medical imaging and radiation therapy

“…Samala et al. first pretrained a DCNN on ImageNet or a larger mammogram dataset and then fine‐tuned on a digital breast tomosynthesis (DBT) dataset for the classification and detection of masses on DBT. Zheng et al .…”