Computer-aided diagnosis system for medical images using deep convolutional neural networks

Non-transitory computer-readable storage media encoded with a computer program including instructions executable by a processor to create an application applying deep convolutional neural networks to medical images to generate a real-time or near real-time diagnosis or diagnostic recommendation, the application comprising:

• a) a software module performing image segmentation of a plurality of medical images, the image segmentation comprising isolating a region of interest from each image;
• b) a software module applying a cascaded deep convolutional neural network detection structure to the segmented images, the detection structure comprising:
• i) a first stage employing a first convolutional neural network to screen all possible locations in each 2D slice of the segmented medical images by a sliding window methodology to identify one or more candidate locations;
• and
• ii) a second stage employing a second convolutional neural network to screen 3D volumes constructed from the candidate locations by selecting at least one random location within each volume with a random scale and a random viewing angle to identify one or more refined locations and classifying the refined locations;
• and
• c) a software module automatically generating a report comprising a diagnosis or diagnostic recommendation.

A computer-implemented method of applying deep convolutional neural networks to medical images to generate a real-time or near real-time diagnosis or diagnostic recommendation comprising:

• a) performing, by a computer, image segmentation of a plurality of medical images, the image segmentation comprising isolating a region of interest from each image;
• b) applying, by the computer, a cascaded deep convolutional neural network detection structure to the segmented images, the detection structure comprising:
• i) a first stage employing a first convolutional neural network to screen all possible locations in each 2D slice of the segmented medical images by a sliding window methodology to identify one or more candidate locations;
• and
• ii) a second stage employing a second convolutional neural network to screen 3D volumes constructed from the candidate locations by selecting at least one random location within each volume with a random scale and a random viewing angle to identify one or more refined locations and classifying the refined locations;
• and
• c) automatically generating, by the computer, a report comprising a diagnosis or diagnostic recommendation.