I currently study aspects of processing and analysis of ultrasonographic images with applications to the study of human and animal reproduction.  Recent directions in this research are summarized below.

Segmentation of Ovarian Structures

The study of ovarian morphology and function often requires identification, enumeration and measurement of the anatomical structures within the ovary, for example, the follicle outlined in red in the image to the left (follicles are fluid-filled cavities that contain developing eggs).  Due to the comparatively low quality of ultrasound images this can be a difficult task.  Automated algorithms are sought which can find the precise outlines of these structures while ignoring other structures that may be present in surrounding tissues.  These so-called “segmentation” algorithms are crucial for automating analyses for higher-level applications which require knowledge of the number, sizes, and arrangements of ovarian structures.

Automated Screening and Computer Assisted Diagnosis

Knowledge about internals of ovaries provided by image segmentation enable further analyses, particularly algorithms which can analyze a scan for signs of disease or dysfunction.  Our lab has developed an automated classification algorithm which correctly classified over 90% of images tested as to whether or not they exhibited morphological abnormalities indicative of polycystic ovary syndrome.  Of further interest are algorithms for detection of ovarian cancer, and what role automated detection algorithms can/should play in the clinical diagnostic process.

Extraction of Physiologic Information

The noisy, or textured appearance of ultrasound images are caused by scattering of the ultrasonic pulses used to acquire the image.  It is hypothesized that this “echotexture” actually contains information indicative of the physiologic status of particular tissue which could, in theory, be recognized and exploited in diagnostic and predictive applications.  It may, for example, be possible to use echotexture to detect which developing ovarian follicle will ovulate, or to accurately guess from the current number, size, and distribution of ovarian follicles within a single image (obtained from segmentation) the current state of the reproductive cycle within 2-3 days, or to gauge the rate of progesterone production of a corpus luteum (another ovarian structure) from it’s echotexture and overall size. 

Digital “Staining”

Classical histologic techniques allow pathologists to visually distinguish different kinds of cells by chemically staining thin sections of biopsy tissue mounted on microscopic slides.  It may be possible to obviate the need for a biopsy by performing an ultrasound scan and digitally “staining” the image with global filtering processes that exploit subtle differences in echotexture to make different kinds of tissue appear more different in ultrasound images.  Indeed, recent work has shown that certain kinds of global filtering improve not only the difference between the visual appearance of the echotexture of luteal (glandular) tissue compared with stroma (vascular and connective tissue) in human ovaries but also the difference between statistical texture descriptors of said echotextures.  This may achieved without knowledge of where each kind of tissue appears in the image (ie. without prior segmentation).