Detecting and Classifying Fetal Brain Abnormalities Using Decision Tree Algorithm
Detecting and Classifying Fetal Brain Abnormalities Using Decision Tree Algorithm
Detecting and Classifying Fetal Brain Abnormalities Using Decision Tree Algorithm
ABSTRACT: Detecting and classifying fetal brain abnormalities from magnetic resonance imaging (MRI) is important, as
approximately 3 in 1000 women are pregnant with a fetal of abnormal brain. Early detection of fetal brain abnormalities
using machine learning techniques can improve the quality of diagnosis and treatment planning. The literature has shown
that most of the work made to classify brain abnormalities in very early age is for preterm infants and neonates not fetal.
However, research papers that studied fetal brain MRI images have mapped these images with the neonates MRI images
to classify an abnormal behaviour in newborns not fetal. In this work, a pipeline process is proposed for fetal brain
classification (FBC) which uses machine learning techniques. The main contribution of this work is the classification of
fetal brain abnormalities in early stage, before the fetal is born. The proposed algorithm is capable of detecting and
classifying a variety of abnormalities from MRI images with a wide range of fetal gestational age (GA) (from 16 to 39
weeks) using a flexible and simple method with low computational cost. The novel proposed method consists of four
phases; pre-processing, segmentation, feature extraction and classification. In the pre-process, the input image is
converted into gray scale and apply the wiener filter for image enhancement. After pre-processing, the adaptive
segmentation is applied for segmenting the image. Then DWT and statistical features are extracted from segmented
image. Finally, the input image is classified using Decision Tree algorithm.
Index Terms - Leaf Classification, Image processing techniques, Morphological operators, Machine Learning algorithms
1. INTRODUCTION
Magnetic resonance imaging (MRI) is an important tool for medical investigation of the brain and is considerably
enhanced in uterus fetal brain imaging. Scanning the fetal brain is essential as approximately 3 in 1000 pregnancies have fetuses
with different types of brain abnormalities. Moreover, several series of neuropathological variations happen, some related to
serious clinical morbidities. Recently, fetal MRI has been considered as a neoteric, non-invasive tool for monitoring fetal brain
development. Fetal MRI confers high contrast resolution for the tissue. Therefore, it confers detailed fetal structural information
of the brain. Machine learning techniques can be used to detect and classify brain abnormalities and tumors from fetal MRI
images at an early stage without any surgical interventions. Early detection of the fetal brain abnormalities will indicate how the
pregnancy will be managed, possible treatments that can be taken, and help parents understand and be prepared for dealing with
the abnormality
Moreover, the early discovery of these abnormalities can improve the quality of diagnosis and follow-up planning.
Machine learning is a well-known field in engineering and computer science. It is a great tool used in pattern recognition and
image processing. It can be applied for medical images to help physicians in rendering medical diagnoses. Also, it prevents
human-based diagnostic error and the effort during an examination. It is can be used as well for fetal brain MRI images for early
detection of abnormalities, diagnosis, and classification of certain diseases. Most of the work that used fetal MRI images focused
on segmentation. However, few works used fetal MRI images for classification. To our knowledge, papers that used fetal brain
images for classification are limited. Although these papers used fetal brain MRI images for classification, they used these images
to predict an abnormal neuro-behavior in small for gestational age (SGA) neonates, not fetuses. This means that they classify the
SGA abnormality after the fetuses are born and become neonates. The authors of these papers assumed the hypothesis that fetal
MRI textural patterns are related to new-born neurobehavioral.
Machine learning is an exciting field of research in computer science and engineering. It is considered a branch of artificial
intelligence because it enables the extraction of meaningful patterns from examples, which is a component of human intelligence.
The appeal of having a computer that performs repetitive and well-defined tasks is clear: computers will perform a given task
consistently and tirelessly; however, this is less true for humans. More recently, machines have demonstrated the capability to
learn and even master tasks that were thought to be too complex for machines, showing that machine learning algorithms are
potentially useful components of computer-aided diagnosis and decision support systems. Even more exciting is the finding that
in some cases, computers seem to be able to “see” patterns that are beyond human perception.
2. LITERATURE REVIEW
[1] They addressed this issue by automatic localization of the fetal anatomy, in particular, the brain which is a structure of
interest for many fetal MRI studies. We first extract superpixels followed by the computation of a histogram of features for each
superpixel using bag of words based on dense scale invariant feature transform (DSIFT) descriptors. We construct a graph of
superpixels and train a random forest classifier to distinguish between brain and non-brain superpixelsIn
[2] the proposed system, at first the self-organising map neural network trains the features extracted from the discrete wavelet
transform blend wavelets and the resultant filter factors are consequently trained by the K-nearest neighbour and the testing
process is also accomplished in two stages. The proposed two-tier classification system classifies the brain tumours in double
training process which gives preferable performance over the traditional classification method. The proposed system has been
validated with the support of real data sets and the experimental results showed enhanced performance
In[3] this study, they apply machine-learning methods to compare whole-brain functional connectivity in preterm infants at term-
equivalent age and healthy term-born neonates in order to test the hypothesis that preterm birth results in specific alterations to
functional connectivity by term-equivalent age.
[4]Machine learning is a technique for recognizing patterns that can be applied to medical images. Although it is a powerful tool
that can help in rendering medical diagnoses, it can be misapplied. Machine learning typically begins with the machine learning
algorithm system computing the image features that are believed to be of importance in making the prediction or diagnosis of
interest. The machine learning algorithm system then identifies the best combination of these image features for classifying the
image or computing some metric for the given image region. There are several methods that can be used, each with different
strengths and weaknesses. There are open-source versions of most of these machine learning methods that make them easy to try
and apply to images. Several metrics for measuring the performance of an algorithm exist; however, one must be aware of the
possible associated pitfalls that can result in misleading metrics. More recently, deep learning has started to be used; this method
has the benefit that it does not require image feature identification and calculation as a first step; rather, features are identified as
part of the learning process. Machine learning has been used in medical imaging and will have a greater influence in the future.
Those working in medical imaging must be aware of how machine learning worworks
[5]Recent resting-state functional MRI investigations have demonstrated that much of the large-scale functional network
architecture supporting motor, sensory and cognitive functions in older pediatric and adult populations is present in term- and
prematurely-born infants. Application of new analytical approaches can help translate the improved understanding of early
functional connectivity provided through these studies into predictive models of neuro developmental outcome. One approach to
achieving this goal is multivariate pattern analysis, a machine-learning, pattern classification approach well-suited for high-
dimensional neuro imaging data. It has previously been adapted to predict brain maturity in children and adolescents using
structural and resting state-functional MRI data.
[6] They present a technique for real-time adaptive thresholding using the integral image of the input. Our technique is an
extension of a previous method. However, our solution is more robust to illumination changes in the image. Additionally, our
method is simple and easy to implement.Our technique is suitable for processing live video streams at a real-time frame-rate,
making it a valuable tool for interactive applications such as augmented reality. Source code is available online.
[7]The purpose of this study was to describe fetal brain anomalies identified during nuchal translucency (NT) examination and
their clinical management. The current report highlights the evolving ability to detect CNS malformation during NT scan. FTS
novel findings expand our embryological understanding of early fetal development. While severe anomalies can be managed by
early termination, other defects necessitate comprehensive evaluation and repetitive imaging.
[8] We tested the hypothesis whether texture analysis (TA) from MR images could identify patterns associated with an abnormal
neurobehavior in small for gestational age (SGA) neonates. Ultra sound .Ultra sound and MRI were performed on 91 SGA fetuses
at 37 weeks of GA. Frontal lobe, basal ganglia, mesencephalon and cerebellum were delineated from fetal MRIs. SGA neonates
underwent NBAS test and were classified asabnormal if $1 area was ,5th centile and as normal if all areas were .5th centile.
Textural features associated with neurodevelopment were selected and machine learning was used to model a predictive
algorithm. Fetal brain MRI textural patterns were associated with neonatal neurodevelopment. Brain MRI TA could be a useful
tool to predict abnormal neurodevelopment in SGA.
[9] In this paper, we present a fully automatic brain tumor segmentation method based on Deep Neural Networks (DNNs). The
proposed networks are tailored to glioblastomas (both low and high grade) pictured in MR images. By their very nature, these
tumors can appear anywhere in the brain and have almost any kind of shape, size, and contrast. These reasons motivate our
exploration of a machine learning solution that exploits a flexible, high capacity DNN while being extremely efficient. Here, we
give a description of different model choices that we’ve found to be necessary for obtaining competitive performance. We explore
in particular different architectures based on Convolutional Neural Networks (CNN), i.e. DNNs specifically adapted to image data
[10] they propose a novel multikernel support vector machine classification framework by using the connectivity features
gathered from WM connectivity networks, which are generated via multiscale regions of interest (ROIs) and multiple diffusion
statistics such as fractional anisotropy, mean diffusivity, and average fiber length
Data Acquisition :
The dataset used in the proposed method consists of 20 fetal images (113 normal and 114 abnormal brain MRI images)
between 16-39 weeks fetal GA. Details on the dataset can be found in [15]. Stacks of T2 weighted images in axial, coronal, and
sagittal planes are obtained with half Fourier single shot RARE technique. Because of the fetal motion throughout the
examination, each acquisition serves as the scout for the subsequent acquisition.
Segmentation Phase:
Firstly, the fetal brain is extracted from the cerebrospinal fluid (CSF) and the amniotic fluid that surrounds the fetal head.
For this step, an optimal intensity threshold T is estimated using adaptive local threshold method [16] .This threshold is used to
separate the pixels of an image into two classes foreground and background. A binary image of the fetal brain Ibinary will be
generated from the original image I after this step using (1):
Next, some morphological operations as shown in fig.1 are applied to remove maternal tissues that are connected weakly
with the outer part of the fetal skull to form a new binary image called Iclose . Afterwards, the Iclose image's boundaries are
traced and classified into parent boundaries and children boundaries where; the parent boundaries are the outer boundary of the
skull and the children boundaries are the holes inside the skull which are completely enclosed by the parents' boundary. The
binary image obtained after defining parent and children boundaries is called Iregion.. Next, Watershed segmentation is applied
on Iregion to remove any other connection from the maternal tissue and skull. A new image is then formed and called
Iwatershed.
Watershed segmentation
Watershed segmentation technique is a region-based segmentation approach. It uses Watershed transform to find out the
"Watershed ridge lines" or "catchment basins" from a binary image by considering it as a surface, where; the black pixels are
considered as the low elevation and the white pixel are the high elevation.
Following this step, the area of all the objects remaining from Iwatershed is measured using region properties method. They are
then sorted in descending order and the largest area is extracted to form a new binary image called IABS. After that, the IABS is
segmented from the original image I using a mask named Imask to produce the final segmented whole fetal brain image Ifinal
using equation (2). Images after these segmentation steps are shown in fig 2.
Firstly in this phase, 4 -decomposition levels of 2D-DWT are applied on the enhanced minor ROI image to extract some
discriminate features. The wavelet basis function used is the discrete approximation of Meyer wavelet (dmey) with mode
periodization. After getting the 4-decomposition levels, a features matrix is generated containing; the approximate, horizontal,
vertical, and diagonal coefficients of the 4th decomposition level.
4) Classification Phase
In this phase, the 13 feature extracted is used to train and test machine learning classifiers. These classifiers include
decision tree. 5-cross validation technique was used for the validation of the classifiers.
DECISION TREE
The core algorithm for building decision trees called ID3 by J. R. Quinlan which employs a top-down, greedy
search through the space of possible branches with no backtracking. ID3 uses Entropy and Information Gain to construct a
decision tree.
Entropy
A decision tree is built top-down from a root node and involves partitioning the data into subsets that contain
instances with similar values (homogenous). ID3 algorithm uses entropy to calculate the homogeneity of a sample. If the
sample is completely homogeneous the entropy is zero and if the sample is an equally divided it has entropy of one.
Information Gain
The information gain is based on the decrease in entropy after a dataset is split on an attribute. Constructing a
decision tree is all about finding attribute that returns the highest information gain (i.e., the most homogeneous branches).
Entropy
E T , X Pc E c
cX
Information Gain
4. RESULTS
5. CONCLUSION
A pipeline method for detecting and predicting the fetal brain abnormalities was proposed. The main contribution of
this paper is the classification of fetal brain abnormalities using a simple and low computational cost method. The proposed
method successfully classified fetal brain abnormalities with images of different gestational age (from 16 weeks to 39 weeks).
Moreover, the technique was capable of predicting different abnormalities not only one type of abnormality. Most of the work
done for classification of MRI images of very young gestational age is for preterm and neonatal. The only work that used fetal
MRI images for classification have mapped these images with their neonatal MRI images to predict small for gestational age
(SGA) abnormality in new-borns not in fetal. Therefore, the proposed method was compared with other methods for classifying
preterm and neonatal. The results showed that the proposed method outperform most of these methods. The results are promising;
however future work will focus on improving the performance of the proposed algorithm and increasing the size of the dataset.
This work will encourage several researchers to start on working on fetal brain classification.
6. REFERENCE
[1] Automatic Brain Localization In Fetal Mri Using Superpixel Graphs [Amir Alansary ,Matthew Lee, Kevin Keraudren,
Bernhard Kainz,Christina Malamateniou,Mary Rutherford,Joseph V. Hajnal,Ben Glocker,Daniel Rueckert 30 December 2015]
[2]BRAIN T UMOUR CLASSIFICATION USING TWO-TIER CLASSIFIER WITH ADAPTIVE SEGMENTATION TECHNIQUE [ V. ANITHA S.
MURUGAVALLI 2016]
[3]MACHINE-LEARNING TO CHARACTERISE NEONATAL FUNCTIONAL CONNECTIVITY IN THE PRETERM BRAIN[ G BALL ET
AL. NEUROIMAGE 2016]
[4] MACHINE LEARNING FOR MEDICAL IMAGING [BRADLEY J.ERICKSON , PANAGIOTIS KORFIATIS, ZEYNETTIN AKKUS, TIMOTHY
L. KLINEFEB 17 2017]
[5] PREDICTION OF BRAIN MATURITY IN INFANTS USING MACHINE-LEARNING ALGORITHMS, [ CHRISTOPHER D. SMYSER, NICO
U.F. DOSENBACH, [...], AND JEFFREY J. NEIL 2016 ]
[6] Adaptive Thresholding Using The Integral Image [ Derek Bradley, Gerhard Roth 30 Jan 2011 ]
[7] Fetal Brain Anomalies Detection During The First Trimester: Expanding The Scope Of Antenatal Sonography[ Eldad
Katorza Et Al. J Matern Fetal Neonatal Med 2018 ]
[8] Automatic Quantitative Mri Texture Analysis In Small-For-Gestational-Age Fetuses Discriminates Abnormal Neonatal
Neurobehavior [ Magdalena Sanz-Cortes Et Al. Plos One. 2013]