A human spine is a complicated structure of bones, joints, ligaments and muscles which all undergo a process of change with age. This paper describes the use of artificial intelligence in visualization and better understanding of the progressive and degenerative changes in human lumbar spine. Visualizing this pattern of change will be helpful in finding the correlations among the spinal features and understanding of how a change in one feature affects others. The self-organizing map (SOM) is an efficient tool for visualization of multidimensional numerical data. It is capable of projecting high-dimensional data onto a regular, usually 2-dimensional grid of neurons. In this paper, SOM is used to visualize the pattern of change in lumbar spine features with the varying age. The paper gives an idea of how the information can be acquired from SOM representations and how the SOM can be best utilized in exploratory data visualization. Data from the lumbar spine MRIs of 61 patients (both male and female) were used in this study. The age of patients ranged from 2 to 93 years. Information for vertebral height, disc height and disc signal intensities were recorded from the MRI scans. SOM then transformed the larger feature space to a smaller one for getting a more meaningful relation between the spinal features. Complexity is reduced and the data set is represented in the form of 2D map which is easier to understand and provides visual description.
A human spine is a complicated and key component of
human being. During the normal ageing process, spine
undergoes progressive and regressive changes which
presumably follow certain pattern. This research focuses
explicitly on the study of progressive and degenerative
changes occurring in the human lumbar spine with the
normal ageing process. This research work concentrates on
the identification and classification of age-related
variations in "human spine" with the help of Magnetic
Resonance Image (MRI). These scans of the lumbar spine
area belong to patients of different age groups. Back pain
is usually associated with the spine disorder. It is the
second most common reason for visits to the doctor’s
clinic, outnumbered only by the upper-respiratory
infections [
A self-organizing map (SOM) is a special type of artificial
neural network which is trained using unsupervised
learning to produce a low-dimensional (typically
twodimensional), discretized representation of the input space
of the training samples, called a map. Unlike to other
artificial neural networks, self-organizing map uses a
neighborhood function to preserve the topological
properties of the input space [
A human spine consists of bones, joints, ligaments and
muscles. There are a total of 33 vertebrae in the human
spine: 7 in the neck (cervical region), 12 in the middle
back (thoracic region), 5 in the lower back (lumbar
region), 5 that are fused to form the sacrum and the 4
coccygeal bones that form the tailbone [
The data set used in this research was taken from the University Hospital Coventry and Warwickshire (UHCW), United Kingdom. The raw data is in the form of Magnetic Resonance Images (MRI) specifically of the lumbar spine area. The format of data is Digital Imaging and Communications in Medicine (DICOM). These magnetic resonance images are the actual scans of the patients. Figure 2, shows the lumbar spine MRI.
Fig 2: Sagittal and an axial view of the lumbar spine MRI Information associated with each MRI scan is the age and gender of the patient which is used for SOM modeling. 3 2 4 3 3 4 6 4 4 2 35 2 MRI scans of 61 patients were selected to develop an initial model. Age and gender distribution of patients are shown in the Table 1 below. Ten groups were formed on the basis of age decades as: G0: up to 10 years, G1: 11-20 years, G2: 21-30, years, G3: 31-40 years, G4: 41-50 years, G5: 51-60 years, G6: 61-70 years, G7: 71-80 years, G8: 81-90 years and G9: 91 and above years of age). There are lots of notable features which can be studied from a lumbar spine MRI scan. The scoring criteria were set to look initially on the vertebral height (L1, L2, L3, L4 and L5), disc height (T12-L1, L1-L2, L2-L3, L3-L4, L4-L5 and L5-S1) and disc signal (T12-L1, L1-L2, L2-L3, L3-L4, L4-L5 and L5-S1). These 17 spinal features were used as an input to build and test the initial model. These features were measured and recorded from the lumbar spine MRIs of 61 patients.
Self-organizing maps (SOMs) are a data visualization technique invented by Teuvo Kohonen which reduces the dimensions of data through the use of self-organizing neural networks. As the humans simply cannot visualize high dimensional data so this technique was created to help us understand high dimensional data. The way SOMs go about reducing dimensions is by producing a map of usually 1 or 2 dimensions which plot the similarities of the data by grouping similar data items together. So SOMs accomplish two things, they reduce dimensions and display similarities. The proposed model has a set of 17 input vectors arranged as columns in a matrix. SOM groups or ranks each sample (patient) on the basis of similarities in their 17 features and assigns certain location to each sample in the map. Figure 3 below; shows the step by step demonstration of the methodology used.
Self-Organizing Map (SOM) is a data visualization
technique which helps to understand high dimensional data
by reducing data dimensions and displaying similarities
among data. According to Teuvo Kohonen; the
selforganizing map (SOM) is a new, effective software tool
for the visualization of high dimensional data. It converts
complex, nonlinear statistical relationships between
highdimensional data items into simple geometric relationships
on a low-dimensional display. As it thereby compresses
information while preserving the most important
topological and metric relationships of the primary data
items on the display, it may also be thought to produce
some kind of abstractions [
SOM contains two processes: training and mapping. In
training process, it constructs the map using input samples.
After the training, it automatically classifiers a new input
sample in the mapping process. The map consists of
several neurons which associated with a weight vector that
has the same dimension as the input sample and a position
in the map. The neurons are arranged originally in physical
positions according to a topology function, such as a grid,
hexagonal, or random topology. The purpose of SOM is to
detect regularities and correlations in their input, and also
to recognize groups of similar input vectors [
Competition: For each input pattern, the neurons compute their respective values of a discriminant function which provides the basis for competition. The particular neuron with the smallest value of the discriminant function is declared the winner.
Cooperation: The winning neuron determines the spatial location of a topological neighbourhood of excited neurons, thereby providing the basis for cooperation among neighbouring neurons.
Adaptation: The excited neurons decrease their individual values of the discriminant function in relation to the input pattern through suitable adjustment of the associated connection weights, such that the response of the winning neuron to the subsequent application of a similar input pattern is enhanced. Unlike other learning technique in neural networks, training a SOM requires no target vector. A SOM learns to classify the training data without any external supervision. Each node's weights are initialized. If the input space is D dimensional (i.e. there are D input units) we can write the input patterns as: And the connection weights between the input units i and the neurons j in the computation layer can be written as: wj = {wji : j = 1, …, N; i = 1, …, D} “N” is the total number of neurons. To determine the best matching unit, one method is to iterate through all the nodes and calculate the Euclidean distance between each node's weight vector and the current input vector. The node with a weight vector closest to the input vector is tagged as the BMU. The Euclidean distance is given as: Where x is the current input vector and w is the node's weight vector.
Network Architecture In SOM, the network is created from a 2D lattice of 'nodes', each of which is fully connected to the input layer. Figure 6 shows a very small Kohonen network of 3 x 3 nodes connected to the input layer shown in dark blue. Each node has a specific topological position (an x, y coordinate in the lattice) and contains a vector of weights of the same dimension as the input vectors. That is to say, if the training data consists of vectors, X, of n dimensions: (x1, x2, x3...xn). Then each node will contain a
The measurements taken from the lumbar MRI of 61
patients were used to model the SOM. Each patient has 17
features which were used as input to the model. These 17
input variables are vertebral heights (5 variables), disc
height (6 variables) and disc signal (6 variables). So the
variables 1-5 are the vertebral height (L1-L5), variables
611 are the disc heights from T12/L1--L5/S1 and variables
12-17 are the disc signals from T12/L1—L5/S1
respectively. The inputs vertebral heights, disc heights and
disc signals have difference ranges. Initial model was built
without normalization of the inputs. Figure 7, below shows
the SOM model built on the basis of 17 input variables
without normalization. In this mode, final quantization
error was: 47.292 and final topographic error was: 0.00.
Figure 8, SOM and U-matrix without normalization of inputs
There are two separate parts of the SOM display. These
include the unified matrix or U-matrix, and the component
planes that are provided for individual variables [
The neurons are drawn into distinct clusters during model
training. Relative distances between neuron clusters are
displayed by the intensity of the colors, with dark color
representing greater distance [
U-matrix Variable5 Variable10 Variable15 d d
The objective of the SOM analysis was to observe interrelationships that exist between 17 variables that were tested and thereby provides a basis for more advance analysis. The SOM does not replace existing statistical tools, but complements our ability to examine relationships between disparate types of variables in a visual presentation of the data. By visualizing the SOM results obtained by normalized dataset, it was concluded that lumbar spine vertebral height does not correlate with the age whereas disc height shows somewhat correlation with age. Disc signal intensities of lumbar spine show a strong correlation with the age. In future, other spinal features will be incorporated to study the spinal aging process in more depth.
This project was partially supported by Warwick Impact Fund, University of Warwick, United Kingdom. Authors would like to thank the University Hospital Coventry and Warwickshire (UHCW) NHS Trust, Coventry, United Kingdom for providing valuable data in the form of magnetic resonance images of the lumbar spine.