An investigation of local energy surface detection integrated with neural n
etwork techniques for image segmentation is presented, as applied in the fe
ature extraction of chromosomes from image datasets obtained using an exper
imental confocal microscope. Use of the confocal microscope enables biologi
sts to observe dividing cells (living or preserved) within a three-dimensio
nal (3-D) volume, that can be visualised from multiple aspects, allowing fo
r increased structural insight. The Nomarski differential interference cont
rast mode used for imaging translucent specimens, such as chromosomes, prod
uces images not suitable for volume rendering. Segmentation of the chromoso
mes from this data is,. thus, necessary.
A neural network based on competitive learning, known as Kohonen's self-org
anizing feature map (SOFM) was used to perform segmentation, using a collec
tion of statistics or features defining the image. Our past investigation s
howed that standard features such as the localized mean and variance of pix
el intensities provided reasonable extraction of objects such as mitotic ch
romosomes, but surface detail was only moderately resolved. In this current
work, a biologically inspired feature known as local energy is investigate
d as an alternative image statistic based on phase congruency in the image.
This, along with different combinations of other image statistics, is appl
ied in a SOFM, producing 3-D images exhibiting vast improvement in the leve
l of detail and clearly isolating the chromosomes from the background.