RT 202 Module 3: Basic medical image processing and analysis

Digital medical imaging

• Image processing and analysis involve a series of steps to enhance and interpret for diagnostic purposes

• x-ray CT scan, MRI or ultrasound where sensors convert the physical signals into digital data

Contrast and density

Photographic properties

Spatial resolution, Distortion, and Magnification

Geometric Properties

Image processing and analysis in film-based radiography

Once the film is developed image analysis techniques includes:

a. Photographic properties = contrast and density

b. Geometric properties = spider resolution, distortion, and magnification

• This accuracy of image analysis depends on both quality of the film and the skill of the radiologist

Image processing and analysis in digital imaging

• Principal advantage: the ability to pre-process and post process the image to extract even more information
  

• offers a more dynamic and automated approach to both pre-processing and processing
  

Pre-medical imaging processing

It involves cleaning and preparing medical images:

a. Removing noise

b. Enhancing contrast

c. Standardizing size and brightness

d. Correcting artifacts

e. Aligning images

COMPONENTS OF PREMEDICAL IMAGE PROCESSING

• Noise reduction

• Image, resizing and normalization

• Artifactory removal

• Image registration (if needed)

• Edge preservation

1. NOISE REDUCTION

• Purpose: to remove unwanted variation(noise) that can obscure important details in the image
  

• Methods: filtering techniques like gaussian, median or mean filters smooth out the image while preserving edges.

Gaussian filtering

 it is a common technique used to reduce image no random variation in brightness or color

GAUSSIAN FILTERING

common for smoothing noise (random variation)

MEDIAN FILTERING

Ideal for removing salt-and-pepper noise while preserving sharp edges

Median Filtering results:

• Sharp Bone Detail

• Close up Detail

• Preserved Soft Tissue
  
  

Key advantage: Exceptional edge-preservation and salt-and-pepper removal

Gaussian Filtering results:

• Smooth soft tissue

• Detail softening

Key Advantage: Effective smoothing of random noise variation can cause slight detail loss

SHARP EDGES

• Preserved = Median

• Softened = Gaussian

NOISE

• All types removed = Median

• Random variation smoothed = Gaussian

2. IMAGE, RESIZING AND NORMALIZATION

• Purpose: to standardize the image, size and brightness for consistency, especially when using automated analysis tools
  

• Methods: resembling or resizing images to a common dimension and intensity, normalization ensures, pixel values fall within a certain range across all images

RESIZING

Step 1: Adjust all images to the same dimensions for standardized input

• Cubic

• Interpolation

NORMALIZATION

Step 2: Rescale pixel values for consistent contrast across all images

• Method A = min-max (original)

• Method B = z-score (transformed)

Pre-processing workflow

Noisy original → Noise reduction → Resizing → Normalization → Ready for Model Training

3. ARTIFACT REMOVAL

• Purpose: to eliminate artifacts caused by patient, metal implants, or scanners that could interfere with diagnosis
  

• Methods: motion correction, algorithms, and specialized software to detect and removes scan artifacts

Artifact Removal: IDENTIFICATION & FILTERING (Step 1)

• Original: Noisy/Artifacted

• Targeted inputs

• Gaussian Filter: Smooths out Gaussian noise overall structure

• Median Filter: Removes salt-and-pepper noise; sharpens edges

• Motion, Streaks, Metal, and Noise

Artifact Removal: IDENTIFICATION & FILTERING (Step 2)

• Motion Blur Removal

• Streak Artifact Removal

• Metal Reduction

• Noise Reduction

Artifact Removal: IDENTIFICATION & FILTERING (Step 3)

Removed: random noise variation, salt-and-pepper noise, blur streaks metal distortion

Preserve: sharp edges, soft tissue detail bone structure

MOTION BLUR REMOVAL

Deblurs to restore sharp detail

STREAK ARTIFACT REMOVAL

Eliminates beam-hardening streaks

METAL REDUCTION

Lessens distortion from metal metal implants

NOISE REDUCTION

Removes residual variations

4. IMAGE REGISTRATION (if needed)

• Purpose: to align images taken different times or from different imaging modalities (like MRI and CT scan)
  

• Methods: Geometric transformations, such as rotation, scaling, or translation

STEP 1: REGISTARTION (ALIGNMENT)

• Translation

• Rotation

• Scaling
  
  

Identify: Motion, Streak, Metal, Noise

Goal: Standardize image positions and ensure point by point correspondence

5. Edge Preservation

• Purpose: To retain important structural boundaries while removing unnecessary details.
  
  

• Methods: Edge-preserving filters like bilateral filters that smooth flat regions while keeping edging sharp

CODING AND DECODING

• It is the process of converting the medical images and data into digital format (?) and then interpreting or retrieving them for viewing and use of (?)

• Allows efficient storage, transmission, and access of medical images across healthcare system

STEP 1: CODING

• Digital conversion & standardizing

• Digital encoder step
  
  a. Basic DICOM coding - Make standard digital files
  b. Adaptive Compression - Efficiently reduce file size

Goal: Converts analog scans and data into standardized digital formats for efficient

STEP 2: DECODING

• Interpretation & Access

• Digital decoder step
  
  c. Secure Retrieval - Secure network access
  d. Interpretation View - Diagnostic Display

Goal: Retrieve and decompress files for viewing, preserving critical anatomical

Data compression and Format Encoding

CODING in PACS?

CODING

is how images are digitally stored, transmitted, or compressed in PACS

DICOM Format

Medical images are encoded into ______ files, which include both images and patient information medical images are encoded into diagram fast, which include both images and patient information

Compression

This helps reduce storage space and speed up transmission this helps reduce storage space and sped up in transmission

Lossless and Lossy

2 types of compression

No image quality lost

Lossless Compression

Some image quality lost but smaller file size

Lossy Compression

Metadata Coding

Along with the patients name, ID, Date, and modality type are coded into the file

LOSSLESS

Description: Reduces file size without losing quality

Use case: Ideal for diagnostic purposes

LOSSY

Description: Reduces size by removing some data

Use case: Used for fast previews or non-critical viewing

Image viewing and retrieval

DECODING in PACS

DECODING

Happens when PACS system or a radiologist’s workstation reads the coded DICOM files and displays the image in a usable format

DICOM viewers

Software that can decode the DICOM format to show high-quality medical images

Decompression

If an image is compressed, decoding involves decompressing it to restore its original form (especially in lossles formats)

Data Extraction

Decoding also includes extracting metadata (like patient info, scan date, etc.) for records or diagnoses

• Data compression

• Format encoding

CODING

• Image Viewing

• Data Retrieval

DECODING

REASONS FOR DATA COMPRESSION IN PACS

1. To save storage space

2. To speed up image transmission

3. To reduce costs

4. To improve workflow efficiency

5. For long-term archiving

FORMAT ENCODING IN PACS

Used to standardize, organize, and securely store medical images and related data so that they can be easily shared, viewed, and understood across different systems and devices

MAIN REASONS FOR FORMAT ENCODING IN PACS

1. Standardization with DICOM format

2. Interoperability between devices

3. Combining image + Patient information in one life

4. Data security and integrity

5. Efficient viewing and processing

POST MEDICAL: IMAGE PROCESSING

• Advanced techniques applied to medical images after they have been acquired, preprocessed, and coded-decoded
  

• This aims to enhance the information extracted from the images for more accurate diagnosis, treatment planning, and monitoring
  

• Post-processing deals with analyzing, quantifying, and interpreting the anatomical or functional data within the images

COMPONENTS OF POST MEDICAL IMAGE PROCESSING

1. IMAGE SEGMENTATION

2. IMAGE REGISTRATION

3. FEATURE EXTRACTION 

4. CONTRAST ENHANCEMENT

5. 3D RECONSTRUCTION AND VISUALIZATION

6. IMAGE FUSION

7. QUANTITATIVE ANALYSIS

8. COMPUTER-AIDED DETECTION/DIAGNOSIS (CAD)

9. IMAGE ANNOTATION AND REPORTING

1. IMAGE SEGMENTATION

• Isolate and identifies the specific anatomical structures origin of interest (tumors, organs, blood vessels)

• Algorithms divide the image into parts based on pixel intensity, shape, or texture

• This helps measure volume, shape and location of abnormalities

2. IMAGE REGISTRATION

• It aligns multiple images from different time points or modalities (like CT scan and MR) for compression
  
  

• Geometric transformations: it is a mathematical operation used to align two or more medical images so they can be compared or combine accurately

3. FEATURE EXTRACTION

• If the attack and quantifies specific characteristics such as (edges, texture, intensity) from segmented areas
  
  

• ___________ is the process of identifying and isolating, specific meaningful patterns or characteristics from medical images that are useful for diagnosis, analysis or further computational processing
  
  

• It works by having mathematical algorithms image regions to obtain meaningful data

4. CONTRAST ENHANCEMENT

• increases the difference between light and dark regions, marking a subtle structures more visible
  
  

• It helps read different differentiate between healthy and abnormal tissues, especially in images like CT scans, MRIs or x-rays where slight differences can be diagnostically important

5. 3-D RECONSTRUCTION AND VISUALIZATION

• It creates 3-D models of the body structures from 2-D image slices such as CT scan or MRI
  
  

• It improves understanding of complex anatomy, surgical planning, and patient communication

6. IMAGE FUSION

• it combines images from different modalities, such as PET with CT into one composite image

• Enhanced diagnostic diagnosis by showing both structure and metabolic activity

• It works by overlaying anatomical and functional information for a comprehensive view

7. QUANTITATIVE ANALYSIS

• converts image features into measurable data such as (rumor size, and blood flow rate) using algorithms to extract numerical values from images
  
  

• It support objective diagnosis and treatment monitoring

8. COMPUTER AIDED DETECTION/DIAGNOSIS (CAD)

• It assist the radiologist by automatically Identifying the potential abnormalities
  
  

• AI and machine learning analyze images to highlight suspicious areas
  
  

• It increases diagnostic accuracy and reduces the oversight and sped up interpretation

9. IMAGE ANNOTATION, AND REPORT REPORTING

• ________ is a process of adding text to an image

• A radiologist marks the area of interest and inputs notes using specialized software

BENIFIT

It enhances communication with other clinicians and maintains comprehensive medical records

BASIC ORDER OF OPERATIONS

1. IMAGE ACQUISITION

2. PRE-PROCESSING

3. IMAGE ENCODING

4. TRANSMISSION TO PACS

5. IMAGE DECODING

6. POST-PROCESSING

IMAGE ACQUISITION

The image is captured using a modality like CT, MRI or X-ray

PRE-PROCESSING

Preparing raw medical images for further analysis by improving their quality and making them easier to interpret

IMAGE ENCODING

before transmission or storage in PACS the images may be encoded (compressed) using standards like JPEG 2000 or run-length encoding bandwidth is a concern

TRANSMISSION TO PACS

The encoded image is sent over the network to PACS

IMAGE DECODING

Once received, the images is decoded (decompressed) so it can be displayed and processed properly

POST-PROCESSING

This includes contrast enhancement, zooming, edge sharpening, 3D reconstruction, filtering or other visualization tool to aid diagnosis

FILTERING AS PRE-PROCESSING

→ it is used to prepare the image for storage, analysis, or diagnosis

GOAL: To improve raw image quality by removing unwanted noise or artifacts

Examples:
→ Gaussian Filter - To smooth out random noise

→ Median filter - To remove salt-and-pepper noise

FILTERING AS POST-PROCESSING

→ GOAL: To enhance the visibility of features for better diagnosis

Examples:
→ Edge-Enhancement to make structures clearer
→ High-pass or sharpening filters to bring out fine details