Life-saving imaging techniques developed at Oxford University

Improvements in echocardiographic sequence and mammogram analysis techniques lead to earlier detection of disease and defects.

Imaging various parts of the body is an established and important method for the diagnosis of diseases such as breast cancer, and is also used extensively for the detection of abnormalities in organs such as the heart. Accurate interpretation, and ultimately correct diagnosis, is dependent on the quality of the images. High quality images, however, can often be extremely difficult to achieve even for experts within a given field. Researchers are constantly striving to improve existing techniques so that better quality images can be produced that will facilitate quicker and more accurate diagnoses. Such improvements have the potential to greatly benefit patients’ prospects by enabling earlier commencement of treatment, thus preserving or even improving patients’ quality of life.

To address these imaging needs, researchers in Oxford’s Department of Engineering Science have developed two new imaging techniques with life saving potential in the medical field. They have achieved important and measurable improvements in image quality that will increase the accuracy of diagnosis of serious diseases and defects.

Echocardiographic Sequence Analysis

Subject movement during capture of an image is a major problem in subsequent diagnosis, as the subject must be tracked as it moves from frame to frame (this movement is known as optical flow or image velocity). Measurement of optical flow can improve the image encoding efficiency, or allow enhancement of the display of the movement of some particular tracked part of the image to assist a clinician attempting to make a diagnosis.

The high noise levels of medical images present many difficulties in image processing. For example, the tracking of cardiac walls in ultrasound images is difficult because of the inherently high level of noise in such images and because of the variation in cardiac motion during the cardiac cycle. Several means of identifying and tracking cardiac walls in echocardiograms have been proposed, but it is a difficult task that requires improvement.

In looking to make these improvements, Oxford researchers have recently developed a method for identifying boundary pixels in echocardiographic sequences or other ultrasound image sequences by utilising phase boundary detection followed by optical flow estimation. New contributions to these basic computer vision processes have resulted in a system that is both fast and robust.

Mammogram Analysis (Microcalcifications)

Early correct diagnosis of breast cancer can mean the difference between life and death for the significant proportion of western women affected by the disease. Small clumps of calcium salts – microcalcifications – are often the earliest signs of breast cancer, and appear in 25% of mammograms. Oxford researchers have developed a new method to identify more reliably these clusters.

Calcifications appear as bright spots or clusters of spots; small clustered whorled calcifications are those most likely to indicate malignancy. The existence of microcalcifications in a mammogram is a clear warning of abnormality. Any program to assist a radiologist detect microcalcifications must miss few, if any, clinically important clusters, but equally must not signal too many false positives. With the increasingly vast number of mammograms to be analysed from screening programmes, automated computer-aided detection methods are a necessity.

Although several methods have been proposed for detecting microcalcification clusters, they have all been limited by faults such as the return of too many false positives. Oxford researchers, however, have recently developed a foveal segmentation method, based on differential local contrast in the image, that will significantly reduce the risk of both false negatives and false positives in identifying calcifications in mammograms.