Applications of convolutional neural networks in x-ray imaging

1. Introduction

Visual perceptions begin at the moment light meets the retina which consists of a sheet of photoreceptors that convert light to electrical signals. These signals are sent via the optic nerve to the primary visual cortex. Signal transmissions of visual perceptions can be emulated in computer science, such as deep neural network algorithms that imitate neural processes. In computer science, neural networks which have each neuron in one layer connected to all neurons in the next layer are fully connected and known as multiplayer perceptrons. Convolutional neural networks (CNNs) are a subset of deep neural networks which can be applied to visual image analysis.

CNNs emulate animal visual processes in which an individual visual-cortex neuron reacts to stimuli only in a restricted region of the entire visual field. This restricted region is known as the receptive field of the neuron, and the receptive fields of nearby neurons partially overlap, resulting the entire visual field completely covered by the receptive fields of all neurons. In applied computation and visual-imagery analysis, CNNs are a class of deep learning neural networks that are trained on existing data and then used to predict outcomes, e.g. voice recognition, facial recognition, language translation, or medical image diagnosis.

Applications of CNNs in mammogram diagnosis are typical examples in automatic image diagnosis.  Many images are still been diagnosed by human experts who rely on sufficient image quality to make confident diagnosis from images. X-ray image quality depends on multiple factors. The amount of differential radiations reaching the image detector have significant effects on the image quality. Within a radiation exposure range, the higher the radiation reaches the image detector, the better is the image quality. Scatter radiation reaching the image detector is one of the main noise factors that degrade the image quality. When the relative amount of scatter radiation to primary radiation is large in the image, the image quality is often severely degraded and interferes with making confident diagnosis.

2. Preliminary findings

Preliminary work has evaluated a CNN algorithm using a simulation image and an abdomeninal radiograph. The CNN algorithm was developed to extract image features which were then used to generate a new radiograph with enhanced image quality.

The simulation image was obtained in Monte Carlo simulation with a PMMA phantom and two disc-shaped objects to imitate a large abdomen and the radiation exposures were set at one-thousandth of a typical radiation exposure required for this abdomen size. The simulation image without applied any image processing failed to show the projections of the two objects. After applying our CNN algorithm, the two projections were clearly visible. This might mean that with a perfect image receptor like the one used in the simulation, radiation exposures could be reduced by thousand folds while the image quality would still be good enough for making confident diagnosis.

The abdominal radiograph of an volunteer adult (body-mass-index 30, height 1.8m) was taken with 75 kVp and 16 mAs without using an anti-scatter grid. In practice, acceptable exposure for such abdomen sizes is about 75 kVp and 40 mAs and with an anti-scatter grid. The original radiograph is not of diagnostic quality. Human experts would have difficulty in making a diagnosis from this radiograph. After enhancing it by the CNN algorithm, the enhanced outcome shows significant improvement in image quality and human experts would be confidently making diagnosis from the enhanced radiograph. This means the CNN algorithm has the potential to reduce radiation exposures to patients, in this example exposure reduction by 60%, while still achieving high-quality useful radiographs.

3. Research project

The aims of this research project are to develop and validate CNN algorithms to assist human experts making confident diagnosis and at the same time reduce radiation exposures to patients hence reduce radiation-exposure-induced cancer risks for patients undergoing x-ray examinations. This project also aims at developing CNN algorithms for automatic image diagnosis. 

3.1 Image quality enhancement using CNNs

Image quality may be determined statistically or perceptually. Statistical image quality assessments are such as contrast-detail analysis, signal-noise-ratio analysis (CNR), figure of merit (FOM) analysis. Perceptual image quality evaluation methods are such as visual grading characteristic (VGC) analysis, visual grading analysis (VGA). CNN algorithms for image quality enhancement can be evaluated using appropriate image evalaution methods. CNN algorithms for image quality enhancement can also be evalauted by a cross-comparison with other image quality enhancement methods, such as the Contrast Limited Adaptive Histogram Equalization (CLAHE), the Brightness Preserving Dynamic Fuzzy Histogram Equalization (BPDFHE), and the Histogram Equalization (HE).

3.2 CNN algorithm for diagnosis

Some applications of CNNs are such as mammogram density clarification, lung disease diagnosis, vessel extraction in angiograms. In automatic image diagnosis, one important criterion is the extraction of image features in terms of mathematic representations. Image features can be clarified into different sets based on mathematic operations. To achieve an accurate diagnosis, a CNN algorthim should have the ability to learn 'skills' to differentiate normal features and abnormal/disease structures, the learning process is just like the training process for a radiologist trainee to become an expert in reading X-ray images.

The CNN algorithm we have developed has excellent advantages in image feature extractions. The CNN algorithm has been tested and shows the ability of enhancing image quality by processing each feature extracted from the original image. We are adding functions to the algorithm for differentiation of features to assist diagnosis.

4. Research candidates

We welcome candidates to join us to develop and validate CNN algorithms for image quality enhancement and diagnosis.