You are on page 1of 8

DSP (Lab)

Submitted To:
Mam Shahinza
Submitted By:
Faizan Sheraz
Class:
BSCE-5
Task No:
CEP
Roll No:
20
Date:
7-12-2023
Statement:
A medical imaging system is being developed to aid in the detection of breast
cancer. The system uses digital signal processing techniques to analyze images
obtained from mammography scans and employs machine learning algorithms
to automatically identify regions of interest that may contain tumors. The
images are first preprocessed to enhance contrast and remove noise, and then
segmented to separate the breast tissue from the background. Next, feature
extraction techniques are used to extract morphological, texture, and density
features from the segmented images, which are used as inputs to the machine
learning algorithms. The machine learning algorithms include supervised
learning algorithms such as support vector machines (SVMs), decision trees,
and random forests, as well as unsupervised learning algorithms such as k-
means clustering and principal component analysis (PCA).

Questions.
1. Trace some of the preprocessing steps used in the medical imaging system
to enhance mammography images and remove noise?

ANSWER :

In medical imaging systems, several preprocessing steps are commonly used to enhance
mammography images and remove noise. Here are some of the typical preprocessing steps:
Image Acquisition: Mammography images are acquired using specialized X-ray machines
designed for breast imaging. The imaging system captures the X-ray transmission data from
the breast, producing a raw digital image.
Image Calibration: Raw mammography images may undergo calibration to correct for
variations in the X-ray machine's response and to ensure consistent image quality.
Calibration involves applying correction factors to the raw data, such as dark current
correction, flat-field correction, and exposure normalization.
Image Alignment: In some cases, mammography images may need alignment to correct for
patient motion or positional inconsistencies. Image alignment techniques can be used to
register multiple images acquired at different times or from different angles to ensure
accurate comparisons and analysis.
Image Denoising: Noise reduction is a crucial step in enhancing mammography images.
Various denoising techniques are employed to reduce the inherent noise caused by X-ray
photon statistics and imaging system electronics. Common denoising methods include
spatial filtering, wavelet denoising, and adaptive filtering algorithms.
Contrast Enhancement: Enhancing the contrast of mammography images is essential for
better visualization of subtle details. Techniques like histogram equalization, contrast
stretching, and adaptive contrast enhancement algorithms can be used to improve the
visibility of structures and abnormalities in the images.
Image Smoothing: Smoothing techniques can be applied to reduce image noise while
preserving important structures. Smoothing filters like Gaussian smoothing or median
filtering can help remove high-frequency noise while preserving edge information.
Artifact Removal: Mammography images can be affected by various artifacts, such as grid
lines, motion artifacts, or breast implant-related artifacts. Specific algorithms are employed
to identify and remove these artifacts, ensuring a clearer and more accurate representation
of the breast tissue.
Image Resampling: In some cases, mammography images may need resampling to adjust the
image size or resolution. Resampling techniques can be applied to interpolate or decimate
the image data while preserving important image characteristics.

2. Explain how the images are segmented to separate the breast tissue from
the background, and why is this step important for detecting tumors?

ANSWER :
In the development of a medical imaging system for breast cancer detection, the segmentation step
plays a crucial role in separating the breast tissue from the background in mammography images.
This step is essential for detecting tumors and facilitating accurate analysis. Let's delve deeper into
the segmentation process and its significance:

Segmentation Techniques:
Thresholding: Thresholding involves setting a pixel intensity threshold to distinguish the breast tissue
from the background. Pixels with intensities above the threshold are considered part of the breast
tissue, while those below are considered background. Adaptive thresholding methods can adaptively
determine the threshold based on local image characteristics.

Region Growing: Region growing algorithms start from a seed point or region within the breast tissue
and iteratively expand the region by incorporating neighboring pixels that meet certain criteria (e.g.,
intensity similarity, gradient magnitude). This process continues until no more pixels satisfying the
criteria can be added, resulting in the complete segmentation of the breast region.

Active Contour Models: Active contour models, also known as snakes, utilize deformable curves or
surfaces that adjust their shape and contour to match the boundaries of the breast tissue. These
models minimize an energy functional, which helps them accurately delineate the breast region.

Machine Learning-based Segmentation: Machine learning techniques, particularly convolutional


neural networks (CNNs), can be trained to automatically segment the breast tissue. By learning from
a large dataset of annotated images, these models can classify pixels or regions as either breast or
background, achieving accurate segmentation.

Importance Of Segmentation For Tumor Detection:


Isolation of Relevant Regions: Segmenting the breast tissue from the background allows subsequent
analysis algorithms to focus specifically on the regions of interest within the breast. By excluding
nonrelevant structures and noise, such as pectoral muscles or artifacts, the analysis can concentrate
on potential tumor locations, improving detection accuracy.

Localization and Characterization of Tumors: Accurate segmentation aids in precisely localizing


tumor regions within the breast tissue. This information is crucial for subsequent steps, including
feature extraction and classification, as it enables accurate assessment of tumor size, shape, and
spatial distribution. Precise characterization of tumors aids in diagnosis, staging, and treatment
planning.

Reduction of False Positives: Segmenting the breast tissue helps reduce false-positive detections by
excluding non-tumor regions and minimizing the impact of noise and artifacts. This enhances the
specificity of the detection algorithm, resulting in more reliable and accurate tumor identification.

Overall, segmentation of breast tissue from mammography images is a vital step in detecting breast
tumors. It isolates the relevant regions, facilitates accurate localization and characterization of
tumors, and reduces false positives. By providing a focused and accurate input to subsequent analysis
steps, segmentation significantly contributes to the effectiveness and reliability of the entire breast
cancer detection system.

3. Show some of the feature extraction techniques used in the medical


imaging system to extract morphological, texture, and density features
from the segmented images.

ANSWER :
Here are some of the feature extraction techniques commonly used in medical imaging systems to
extract morphological, texture, and density features from segmented images for breast cancer
detection:

Morphological Features:
Area: Calculating the total number of pixels in the segmented region.

Perimeter: Measuring the boundary length of the segmented region.

Compactness: Calculating the ratio of perimeter squared to area, indicating the compactness of the
region.

Eccentricity: Measuring the elongation or roundness of the segmented region.

Solidity: Calculating the ratio of the region's area to the area of its convex hull, indicating how solid
the region is.
Texture Features:
Gray-Level Co-occurrence Matrix (GLCM): Analyzing the spatial relationships between pixel
intensities in the segmented region to extract features like contrast, energy, homogeneity, and
correlation.

Gray-Level Run Length Matrix (GLRLM): Analyzing the lengths and frequencies of consecutive pixels
with the same intensity level to extract features like run length, gray-level non-uniformity, and run
entropy.

Local Binary Patterns (LBP): Encoding the local texture patterns by comparing the intensity values of
a pixel with its neighboring pixels, extracting features like LBP histogram, variance, and uniformity.

Gabor Filters: Applying a set of Gabor filters with different orientations and frequencies to capture
texture information at various scales and directions.

Density Features:
Density Distribution: Computing the pixel intensity histogram within the segmented region to extract
features like mean intensity, standard deviation, skewness, and kurtosis.

Histogram of Oriented Gradients (HOG): Computing gradient orientations and magnitudes within the
segmented region to extract features representing the local density of oriented gradients.

These feature extraction techniques help capture relevant information from the segmented breast
regions, enabling the machine learning algorithms to learn discriminative patterns associated with
tumor presence. The extracted morphological, texture, and density features serve as input to the
supervised and unsupervised machine learning algorithms, such as SVMs, decision trees, random
forests, k-means clustering, and PCA, which can further analyze the data and classify regions of
interest as potentially cancerous or non-cancerous.

It's important to note that the choice and combination of specific feature extraction techniques may
vary based on the specific goals, available data, and the characteristics of the imaging system.
Additionally, advancements in medical imaging and machine learning continue to introduce new
techniques and approaches to improve the accuracy and efficiency of breast cancer detection.

4. Extracted features used as inputs to the machine learning algorithms, and


shows some examples of the supervised and unsupervised learning
algorithms used in the system?

ANSWER :
Once the features are extracted from the segmented images, they are used as inputs to machine
learning algorithms for breast cancer detection. Here are some examples of supervised and
unsupervised learning algorithms commonly used in the system:

Supervised Learning Algorithms:


Support Vector Machines (SVM): SVMs are powerful classifiers that aim to find an optimal
hyperplane to separate different classes based on the extracted features. They can handle high-
dimensional feature spaces and are effective in handling binary classification tasks.
Decision Trees: Decision trees use a hierarchical structure of branching nodes to classify data based
on a series of attribute-value tests. They are interpretable, easy to understand, and can handle both
binary and multi-class classification problems.

Random Forests: Random forests are an ensemble learning method that combines multiple decision
trees. They reduce overfitting and improve accuracy by averaging predictions from different decision
trees.

Artificial Neural Networks (ANN): ANNs are computational models inspired by the structure and
function of the human brain. They consist of interconnected nodes (neurons) and layers and can
learn complex patterns. Deep learning architectures, such as convolutional neural networks (CNNs),
are commonly used for image classification tasks.

he supervised learning algorithms have been used in this study to distinguish the benign from
malignant breast cancer hence their combination has beeen used to accurately predict the cancer
diagnosis labels.

Here below is the summarized workflow of supervised machine learning algorithms employed.

Workflow of Employed Machine Learning Algorithms

Unsupervised Learning Algorithms:


K-means Clustering: K-means is an iterative algorithm that aims to partition data into K clusters
based on similarity. It can be used to identify patterns or groupings in the extracted features without
prior knowledge of the classes.
Hierarchical Clustering: Hierarchical clustering builds a hierarchical tree-like structure of clusters. It
allows for exploration at different levels of granularity and can reveal inherent subgroups or
structures in the data.

Principal Component Analysis (PCA): PCA is a dimensionality reduction technique that transforms
high-dimensional data into a lower-dimensional space while retaining most of the variance. It can be
used to reduce feature dimensionality and extract the most informative features.

Self-Organizing Maps (SOM): SOM is a type of artificial neural network that performs unsupervised
learning by mapping high-dimensional data onto a lower-dimensional grid. It can reveal the
topological relationships and cluster patterns in the feature space.

These algorithms utilize the extracted features as input to train models that can classify and
distinguish between cancerous and non-cancerous regions in mammography images. The choice of
the specific algorithm(s) depends on factors such as the size and nature of the dataset,
computational resources, and the desired performance and interpretability of the system.

It's important to note that the field of machine learning is vast, and advancements in the
development of new algorithms and architectures are ongoing. The choice of algorithms may vary
based on the specific needs and advancements in the medical imaging field

5. States how is the performance of the medical imaging system evaluated,


and what are some metrics used to assess the accuracy of the system in
detecting breast cancer?

ANSWER :
The performance of a medical imaging system for detecting breast cancer is typically evaluated using
various metrics to assess its accuracy and effectiveness. Here are some common metrics used to
evaluate the performance of such systems:

Sensitivity (True Positive Rate): Sensitivity measures the proportion of actual positive cases
(cancerous regions) correctly identified by the system. It indicates how well the system can detect
breast cancer.

Specificity (True Negative Rate): Specificity measures the proportion of actual negative cases
(noncancerous regions) correctly identified as negative by the system. It indicates the system's ability
to correctly identify regions without breast cancer.

Accuracy: Accuracy represents the overall correctness of the system's predictions, considering both
true positive and true negative cases. It calculates the proportion of correctly classified cases
(cancerous and non-cancerous) out of the total cases.

Precision (Positive Predictive Value): Precision measures the proportion of true positive cases out of
the total cases predicted as positive by the system. It indicates the system's ability to minimize
falsepositive results.
Recall (Sensitivity or True Positive Rate): Recall is another term for sensitivity and measures the
proportion of actual positive cases correctly identified by the system. It focuses on the system's
ability to capture all positive cases.

F1 Score: The F1 score is the harmonic mean of precision and recall. It provides a balanced measure
of the system's performance, considering both false positives and false negatives.

Receiver Operating Characteristic (ROC) Curve: The ROC curve is a graphical representation of the
true positive rate (sensitivity) against the false positive rate (1 - specificity) at various classification
thresholds. It provides a visual assessment of the system's performance across different threshold
settings.

Area Under the ROC Curve (AUC): AUC quantifies the overall performance of the system by
measuring the area under the ROC curve. It provides a single metric that represents the system's
ability to discriminate between cancerous and non-cancerous regions.

These metrics allow for a comprehensive evaluation of the medical imaging system's performance in
detecting breast cancer. The choice of metrics depends on the specific requirements, goals, and
characteristics of the system. It's important to consider a combination of these metrics to gain a
holistic understanding of the system's accuracy, sensitivity, specificity, precision, and overall
performance in breast cancer detection.

COMPLETE PROCESS|:

You might also like