Generalized Kohonen’s competitive learning algorithms for
ophthalmological MR image segmentation
Karen Chia-Ren Lin
a
, Miin-Shen Yang
b,
*, Hsiu-Chih Liu
c
, Jiing-Feng Lirng
d
,
Pei-Ning Wang
c
a
Department of Management Information System, Nanya Institute of Technology, Chung-Li, Taiwan
b
Department of Applied Mathematics, Chung Yuan Christian University, Chung-Li, Taiwan
c
Neurological Institute, National Yang-Ming University and Taipei Veterans General Hospital, Taipei, Taiwan
d
Department of Radiology, National Yang-Ming University and Taipei Veterans General Hospital, Taipei, Taiwan
Received 4 January 2003; received in revised form 25 April 2003; accepted 26 April 2003
Abstract
Kohonen’s self-organizing map is a two-layer feedforward competitive learning network. It has been used as a competitive learning
clustering algorithm. In this paper, we generalize Kohonen’s competitive learning (KCL) algorithm with fuzzy and fuzzy-soft types called
fuzzy KCL (FKCL) and fuzzy-soft KCL (FSKCL). These generalized KCL algorithms fuse the competitive learning with soft competition
and fuzzy c-means (FCM) membership functions. We then apply these generalized KCLs to MRI and MRA ophthalmological segmenta-
tions. These KCL-based MRI segmentation techniques are useful in reducing medical image noise effects using a learning mechanism. They
may be particularly helpful in clinical diagnosis. Two real cases with MR image data recommended by an ophthalmologist are examined.
First case is a patient with Retinoblastoma in her left eye, an inborn malignant neoplasm of the retina frequently metastasis beyond the
lacrimal cribrosa. The second case is a patient with complete left side oculomotor palsy immediately after a motor vehicle accident. Her
brain MRI with MRA, skull routine, orbital CT, and cerebral angiography did not reveal brainstem lesions, skull fractures, or vascular
anomalies. These generalized KCL algorithms were used in segmenting the ophthalmological MRIs. KCL, FKCL and FSKCL comparisons
are made. Overall, the FSKCL algorithm is recommended for use in MR image segmentation as an aid to small lesion diagnosis. © 2003
Elsevier Inc. All rights reserved.
Keywords: Kohonen’s competitive learning network; Fuzzy clustering; Fuzzy-soft method; Magnetic resonance image (MRI); Image segmentation
1. Introduction
The medical image segmentation techniques provide im-
portant diagnostic tools for the clinician. In general, medical
images are obtained using different acquisition methods,
including x-ray computer tomography (CT), single photon
emission tomography (SPET), positron emission tomogra-
phy (PET), ultra-sound (US), magnetic resonance image
(MRI) and magnetic resonance angiographies (MRA), etc.
MRI systems are important in medical image analysis. MRI
has the multidimensional nature of data provided from ei-
ther one of two different pulse sequences. The first pulse
sequence is a spin echo pulse sequence that uses a 90 degree
excitation pulse followed by one or more 180 degree re-
phrasing pulses to generate one or more spin echoes. The
second pulse sequence is a gradient echo pulse sequence
that utilizes RF excitation pulse flips through any angle (not
just 90 degree). The gradient rephrases the magnetic mo-
ments so that a signal can be received by a coil and called
a gradient echo. The timing parameters for the pulse se-
quence, i.e., repetitive time (TR), echo time (TE) and RF
excitation pulse provide sequence information from three
types of tissue-dependent parameters; T1 (the spin-lattice
relaxation time), T2 (the spin-spin relaxation) and PD (the
proton density). However, there is always some degree of
T2 weighting presented on any image due to the absence of
a 180 degree rephrasing pulse. The gradient echo sequence
allows for a reduction in the scan time as the TR is greatly
reduced. The gradient echo sequence may be more useful
than a spin echo pulse sequence to produce dynamic con-
* Corresponding author. Tel.: +886-3-2653100; fax: +886-3-
2653160.
E-mail address: msyang@math.cycu.edu.tw (M.-S. Yang).
Magnetic Resonance Imaging 21 (2003) 863– 870
0730-725X/03/$ – see front matter © 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0730-725X(03)00185-1