Digital Radiography: Image Processing, Spatial Resolution, and Display

Image Processing Stages & Determinants of Brightness and Contrast

  • The formation of a visible digital radiographic image requires data to proceed through two fundamental processing stages:
    • The histogram stage/phase.
    • The lookup table (LUT\text{LUT}) stage.
  • Algorithmic vs. Graphical Nature:
    • Histograms can be visually graphed as a diagram (specifically a bar graph).
    • Lookup tables cannot be visually rendered directly; they consist entirely of mathematical algorithms, binary coding (11s and 00s), and formulas.
  • Governing Factors for Brightness and Contrast:
    • Both the histogram and the lookup table jointly determine image brightness and image contrast.
    • On standardized examinations or clinical assessments, any inquiry regarding what controls or affects image brightness must cite both the histogram and the lookup table.
    • Any inquiry regarding what controls or affects image contrast must cite both the histogram and the lookup table.

Raw Data, Collimation Detection, and Regions of Interest

  • Nature of Raw Image Data:
    • Raw data encompasses every reading collected by the Image Receptor (IR\text{IR}) from the patient exposure prior to computer processing.
    • It contains all recorded values located inside the collimation margins as well as all scatter, background, and off-focus radiation located outside the collimation margins.
    • Unprocessed raw images occasionally flash briefly on an acquisition monitor; they appear distorted, unoptimized, and extremely low in diagnostic quality.
  • Collimation Detection:
    • Processing algorithms search the raw receptor data to locate the collimated exposure borders (e.g., rectangular or square boundaries).
    • Values falling inside the detected collimation borders are retained to generate the image.
    • Values falling outside the collimated margins are subtracted and deleted.
  • Values of Interest (VOI\text{VOI}):
    • The pixel data points retained inside the collimation borders are designated as the Values of Interest (VOI\text{VOI}).
    • Each individual pixel is assigned a specific numerical value representing a discrete shade of gray (e.g., a numerical value of 25002500 represents one gray shade, whereas a numerical value of 5050 represents a distinctly different gray shade).
  • Region of Interest (ROI\text{ROI}):
    • The ROI\text{ROI} is the specific anatomical region from which diagnostic data is acquired once collimation boundaries are identified.
    • If processing software fails to locate the collimation borders, it errantly incorporates all external, unattenuated scatter and raw background exposure into the ROI\text{ROI}.
    • Inclusion of extraneous data skews the histogram calculations, rendering the initial visible image severely compromised or unreadable.
    • Applying a manual collimation box over the desired anatomical area forces system reprocessing; this excludes the erroneous outside data, correctly resets the ROI\text{ROI}, and normalizes the image.

Histogram Formation and Rescaling

  • Histogram Structure:
    • A histogram is a discrete bar graph whose sharp steps have been smoothed into a continuous curve.
    • The horizontal axis (xx-axis) represents the scale of contrast, organized by pixel density values ranging from clear/minimum density (00) to direct exposure/maximum density (20482048).
    • Intermediate pixel values correspond to varying shades of gray.
    • As an example, direct X-ray exposure passing between fingers during a hand examination generates maximum-density pixel values reaching 20482048.
    • The vertical axis (yy-axis) plots the frequency (the total pixel count) for each specific numerical value.
    • Histograms vary dynamically and are unique to every single patient exposure.
  • Histogram Rescaling Process:
    • Image processing units possess preprogrammed anatomical reference histograms detailing ideal graphic distributions (e.g., the expected curve of a standard Anteroposterior (AP\text{AP}) Hip).
    • The system plots the raw data histogram (VOI\text{VOI}) and compares it against the preprogrammed anatomical histogram.
    • The computer reconciles differences between the acquired raw histogram and the reference histogram by mathematically blending them together.
    • This blending process produces a rescaled histogram, generating a rescaled image with stabilized, preliminary brightness and contrast.

Lookup Table (LUT\text{LUT}) Operation

  • Purpose of the Lookup Table:
    • The rescaled image requires further refinement to optimize visible contrast and separate subtle anatomical tissue densities.
    • The LUT\text{LUT} serves as a mathematical formula or transfer function that re-maps input pixel values from the rescaled image into new output pixel values for the displayed image.
  • Mechanism of Value Conversion:
    • If a lookup table operated with a strictly linear 1:11:1 ratio (a straight diagonal line), input pixel values would yield identical output values (e.g., an input of 10241024 outputs as 10241024), leaving contrast and brightness completely unchanged.
    • By altering the algorithm to follow an S-shaped or curved characteristic function, the brightness and contrast are dynamically tailored.
    • For example, an input value of 10241024 mapped along a custom curve may be converted to an output value of 850850, altering the displayed brightness.
    • In a clinical lookup table conversion:
    • An input value of 4545 is converted to an output value of 1111.
    • An input value of 600600 is converted to an output value of 725725.
    • An input value of 14201420 is converted to an output value of 20112011.
    • This non-linear reassignment expands the grayscale separations across the anatomical structure, converting a gray, flat image into a high-contrast diagnostic radiograph.

Exposure Technique, the 15% Rule, and Patient Dose

  • Exposure Latitude in Digital Radiography:
    • Digital systems correct for variations in exposure through histogram rescaling and LUT\text{LUT} mapping.
    • Increasing peak kilovoltage (kVpkVp) by 15%15\% allows a technician or an Automatic Exposure Control (AEC\text{AEC}) system to decrease the milliampere-seconds (mAsmAs) by 50%50\%.
    • Applying this 15%15\% rule reduces patient exposure dose by approximately one-third (30%30\% to 33.3ˉ%33.\bar{3}\%).
    • While patient dose decreases by a third, overall receptor exposure remains stable.
    • The 15%15\% increase can safely be applied once across standard examinations; applying it twice consecutively risks excessive loss of subject contrast that digital algorithms cannot salvage.

Spatial Resolution Governing Factors

  • Definition of Spatial Resolution:
    • The sharpness or structural detail recorded in an anatomical image, governed by the ability to distinguish small adjacent structures.
  • Primary Controlling Factors:
    • Matrix size.
    • Pixel size.
    • Field of View (FOV\text{FOV}).
    • Detector element (DEL\text{DEL}) size (specifically in Digital Radiography (DR\text{DR}) panels).
  • Sampling frequency historically influenced Computed Radiography (CR\text{CR}) spatial resolution.

Matrix, Pixels, and Voxels

  • Matrix Architecture:
    • A digital image matrix is configured as a two-dimensional checkerboard arrangement of columns and rows.
    • The intersections of these columns and rows form picture elements (pixels).
  • Pixels (Picture Elements):
    • Pixels are generally square in geometry (though theoretical models can be circular).
    • Every individual pixel in the matrix is assigned a discrete numerical value corresponding to an assigned shade of gray.
    • Assembled in proper anatomical sequence, pixels form the diagnostic image.
  • Voxels (Volume Elements):
    • A voxel represents a three-dimensional volume of tissue, incorporating slice thickness in Computed Tomography (CT\text{CT}).
    • In CT\text{CT} protocols, scan parameters dictate slice thickness and slice intervals (skipping intervening slices).
    • Missing or overly thick slices run the risk of skipping small anatomical pathologies, such as kidney stones.
    • Scanning every consecutive millimeter without gaps doubles the patient radiation dose, necessitating strict clinical imaging protocols tailored by anatomical region (e.g., specific kidney stone protocols vs. generalized abdomen protocols).
    • Projection radiography deals strictly in two-dimensional pixels; on general radiography examinations, the term voxel represents three-dimensional modalities (CT\text{CT}) and functions as an examination distractor.

Pixel Pitch and Image Sharpness

  • Definition of Pixel Pitch:
    • Pixel pitch is defined strictly as the physical distance measured from the center of one pixel to the center of an adjacent pixel.
  • Impact on Spatial Resolution:
    • A smaller pixel pitch denotes smaller physical pixels, allowing more pixels to be packed into a fixed matrix space.
    • Decreasing pixel size and decreasing pixel pitch directly increases spatial resolution.
    • High spatial resolution permits higher magnification of an image (such as zooming in on fine cortical bone markings) before pixelation and image degradation occur.
    • While smaller pixels provide superior detail, they present technical tradeoffs including increased manufacturing expense and higher data handling demands.

Field of View (FOV\text{FOV}) Variations and Receptor Sizing

  • Field of View (FOV\text{FOV}):
    • The visible anatomical area selected through physical collimation or software cropping.
  • Manufacturer Engineering Options for Receptor Sizing:
    • When comparing large image receptors (14Ă—17 inches14 \times 17\,\text{inches}) to small image receptors (10Ă—12 inches10 \times 12\,\text{inches}), manufacturers engineer the relationship in one of two ways:
    • Method 1 (Fixed Matrix Compression): The full matrix of the larger plate is compressed into the smaller physical dimensions of the 10Ă—12 inches10 \times 12\,\text{inches} cassette. Because the same number of pixels is compressed into a smaller area, the pixel pitch decreases, and the smaller receptor yields higher spatial resolution.
    • Method 2 (Proportional Matrix Truncation): The pixel pitch remains identical between both sizes, and the smaller cassette simply utilizes a proportionally smaller matrix (fewer total pixels). In this configuration, spatial resolution between the large and small plates is identical.

Detector Element (DEL\text{DEL}) Size and DR Resolution

  • Microscopic Architecture of a DEL\text{DEL}:
    • In direct digital radiography panels, each individual pixel corresponds to a hardware detector element (DEL\text{DEL}).
    • Every microscopic DEL\text{DEL} incorporates three primary structural components:
    • A Thin-Film Transistor (TFT\text{TFT}).
    • A storage capacitor.
    • Sensing and data extraction circuitry.
  • Size Relationships and Resolution Values:
    • As physical DEL\text{DEL} sizes decrease, spatial resolution increases, while manufacturing costs rise significantly.
    • A DEL\text{DEL} size of 200 μm200\,\mu\text{m} provides a spatial resolution of 2.5 line pairs per millimeter (lp/mm)2.5\,\text{line pairs per millimeter (lp/mm)}.
    • A DEL\text{DEL} size of 100 μm100\,\mu\text{m} provides a spatial resolution of 5 line pairs per millimeter (lp/mm)5\,\text{line pairs per millimeter (lp/mm)}.
    • Halving the detector element size (200 μm→100 μm200\,\mu\text{m} \rightarrow 100\,\mu\text{m}) doubles the spatial resolution (2.5 lp/mm→5 lp/mm2.5\,\text{lp/mm} \rightarrow 5\,\text{lp/mm}).
  • Direct Capture Radiography:
    • Direct capture DR\text{DR} yields the highest spatial resolution among digital modalities because X-ray energy is converted directly into electrical charges without an intermediate conversion to light.

Digital Radiography vs. Film-Screen Radiography

  • Spatial Resolution Comparisons:
    • Direct capture digital systems achieve approximately 5 lp/mm5\,\text{lp/mm}.
    • Traditional film-screen systems routinely achieved 10 lp/mm10\,\text{lp/mm} or higher, delivering more than double the spatial resolution of modern digital systems.
  • Data Point Capacities:
    • A typical digital matrix of 3000Ă—30003000 \times 3000 pixels contains approximately 9 000 0009\,000\,000 discrete data points.
    • Film-screen emulsion containing microscopic silver halide crystals yielded trillions of structural data points.
    • Thicker film emulsions allowed denser crystal packing, multiplying the recorded data points.
  • Dynamic Processing Capabilities:
    • Film-screen radiography was limited by a static display; once exposed and chemically processed, optical density could not be adjusted.
    • Digital radiography's clinical superiority lies not in spatial resolution, but in wide dynamic range and post-processing capabilities, allowing users to dynamically window, level, and manipulate grayscale values across a wide exposure spectrum.

Modulation Transfer Function (MTF\text{MTF}) and Spatial Frequency

  • Definition and Measurement:
    • Modulation Transfer Function (MTF\text{MTF}) evaluates the physical detector's capacity to preserve spatial resolution or image sharpness.
    • It measures the system's ability to record small anatomical structures in fine detail.
    • System components evaluated include the detector, algorithmic processing, and matrix sizing working synchronously.
  • Association with Spatial Frequency:
    • Spatial frequency is expressed in line pairs per millimeter (lp/mm\text{lp/mm}).
    • In imaging physics assessments, spatial frequency correlates directly with MTF\text{MTF}.
  • Quantitative Scale:
    • MTF\text{MTF} is quantified on a numerical scale from 00 to 1.01.0 (or 0%0\% to 100%100\%).
    • A score of 1.01.0 (or 100%100\%) represents a theoretically perfect reproduction of anatomical detail.
    • When comparing performance values:
    • Given system percentages of 70%70\%, 75%75\%, 60%60\%, and 85%85\%, the system exhibiting 85%85\% possesses the highest MTF\text{MTF}.
    • Given system decimal ratios of 0.700.70, 0.750.75, 0.600.60, and 0.850.85, the score of 0.850.85 demonstrates the greatest structural detail.
  • Technologist vs. Hardware Control of Detail:
    • Hardware MTF\text{MTF} values define the upper limit of image detail under fixed machine specifications.
    • Technologist-controlled factors that maximize physical detail include selecting a small focal spot size, maximizing Source-to-Image Distance (SID\text{SID}), and minimizing Object-to-Image Distance (OID\text{OID}).

Quantization, Pixel Bit Depth, and Grayscale

  • Quantization:
    • The digital processing function performed by the central processing unit (CPU\text{CPU}) wherein an electrical signal measured from an X-ray interaction is assigned a discrete numerical value corresponding to a specific shade of gray.
  • Pixel Bit Depth:
    • The total pool or range of numbers available for the computer to assign to a pixel, determined by the binary numbering formula (2n2^n, where nn equals bit depth).
    • Grayscale is the complete range of different gray shades made available by the bit depth.
  • Visual Limitations vs. Digital Depth:
    • The human visual system can discern approximately 100100 discrete shades of gray (with historical literature occasionally citing as few as 5050 shades).
    • Modern digital systems capture up to 16-bit16\text{-bit} depth, corresponding to 65 53665\,536 available gray shades (216=65 5362^{16} = 65\,536).
    • Binary depth ranges include:
    • Low-range displays: 44 shades of gray.
    • Intermediate displays: 256256 shades of gray (28=2562^8 = 256).
    • High-tier displays: 65 53665\,536 shades of gray (216=65 5362^{16} = 65\,536).
  • Windowing and Leveling Mechanism:
    • Windowing and leveling functions slide the human eye's visible window of 100100 discernible shades up and down along the total spectrum of 65 53665\,536 recorded values.
    • In CT\text{CT} neuroimaging, window adjustments allow dense cranial bone to saturate to white while brain soft tissue is visualized; adjusting the window/level in reverse visualizes bone trabeculae while brain tissue drops out of view.

Modality Bit Depths, Matrices, and Storage Limits

  • Modality Matrix and Bit Depth Comparisons:
    • Digital Radiography (DR\text{DR}): Matrix typically 3000Ă—30003000 \times 3000, operating at a 16-bit16\text{-bit} depth (65 53665\,536 shades).
    • Magnetic Resonance Imaging (MRI\text{MRI}): Commonly uses a 256Ă—256256 \times 256 matrix with a 16-bit16\text{-bit} depth.
    • Computed Tomography (CT\text{CT}): Commonly uses a 512Ă—512512 \times 512 matrix with a 12-bit12\text{-bit} depth (40964096 shades).
    • Digital Mammography: Utilizes the highest matrix size across digital radiography; however, its bit depth is constrained to 12-bit12\text{-bit} to manage total file size.
  • Digital Mammography Resolution Demands:
    • Mammography was the final radiologic discipline to transition from film to digital systems because microcalcifications and early-stage tumors require exceptionally high spatial resolution.
    • If spatial resolution is coarser than the targeted micro-calcification or tumor margin, edge blur (penumbra) dominates the structure, rendering small malignancies invisible.
  • Digital Storage and Server Limitations:
    • Utilizing maximum bit depth (16-bit16\text{-bit}) alongside massive matrix sizes yields extremely large image files.
    • Large file volumes strain network bandwidth, physical and cloud Picture Archiving and Communication System (PACS\text{PACS}) servers, and portable patient distribution media (such as compact discs).

Questions & Discussion

  • Question regarding the definition of quantization:
    • Query: Clarification was requested on whether quantization is defined as the process of assigning a number to a pixel.
    • Response: Quantization is confirmed as the exact processing action of measuring the signal and assigning a numerical value to a pixel. Bit depth is the total pool of numbers available to choose from, and grayscale is the resulting dynamic range.
  • Question regarding the meaning of the "detector" in MTF measurements:
    • Query: Clarification was requested on whether the term detector in MTF\text{MTF} refers to the entire imaging system.
    • Response: In the context of Modulation Transfer Function, the measurement evaluates the holistic system—including the physical detector panel, proprietary reconstruction algorithms, and matrix architecture working together to preserve spatial detail.
  • Question regarding corrective actions for low MTF images:
    • Query: If an image appears degraded due to low MTF\text{MTF}, can a technologist physically correct it on that machine?
    • Response: An MTF\text{MTF} limitation is a machine/hardware threshold. If all technical factors are equal, the machine cannot exceed its built-in MTF\text{MTF}. Technologists can optimize physical resolution factors (selecting small focal spot, increasing SID\text{SID}, decreasing OID\text{OID}), but the underlying hardware and algorithmic processing ceiling remains fixed.
  • Discussion on clinical use of small (10Ă—12 inches10 \times 12\,\text{inches}) digital panels:
    • Countryside Neonatal Intensive Care Unit (NICU\text{NICU}): Technologists utilize small 10Ă—12 inches10 \times 12\,\text{inches} detector plates by sliding them into dedicated under-bed cassette trays to radiograph premature infants.
    • Emergency Departments (Morton Plant and Saint Anthony's): Dedicated small plates are deployed when patients exhibit severe mobility restrictions that prevent positioning on standard 14Ă—17 inches14 \times 17\,\text{inches} panels.
  • Discussion on instructional materials and ADA compliance:
    • Academic course files (such as PowerPoints and PDF syllabi) require structural adjustments to fulfill compliance standards, including unique slide titles without duplication, text descriptions for all graphic elements, and restrictions on highlighted text.
    • Lecture materials and procedure course notes are transitioned to centralized server distribution with direct student login credentials to preserve unrestricted file access across the program duration.