To Do List 1. Content for microCT to BMD training 2. 3. 4. 5. 6. 7. for symbol codes, see http://www.miswebdesign.com/resources/articles/xhtml-symbol-character-references.html Note on
(preformatted) tag: Preserves spaces and line breaks but may be limited to fixed width font (e.g. Courier) Seems to be able to support on events, On DTD rules = STRICT Supports class, id, style, and the following events onclick, ondblclick, etc.. Notes on AAPM TG-18 SECTION ONE - OVERVIEW OF HISTORY OF STANDARDS --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- y2000 NEMA-DICOM Standard (PS 3) * DICOM set by ACR and NEMA * GSDF (Grayscale Standard Display Function NEMA 2000 *GSDF = images Xfered via DICOM be displayed on any DICOM-comp. device with consistent grayscale. *GSDF = equal changes in Dx values caused equal changes in brightness through perceptual linearization *GSDR distinguishes between image optimization via LUT/DICOM and output device standardization Process in general * pixel values/grayscale values: saved to disk after acquistion and device level corrections (12 to 16 bits per image) * p values: presentation values sent to display after application processing (edge enhancement) and proprietary window/level * DDL or digital driving levels: after p-values are converted by display hardware via LUT * Analog monitors: luminence values fron DDL produced by DAC conversion by the display hardware *DDL to luminance Xformation (monitor characteristic function) is usually not adjustabl * LCD monitors: work directly from DDL??? *DICOM standar allows the calculation of a function to map p-values to DDL so that equal changes in perceived brightness correspond to equal changes in p-values. In practice, the char.function is determined by initally applying a unit transformation at the LUT that allows software manipulation of the DDL and direct meas. of the monitor char. function. This function is used to calculate LUT such that the net xformation from [-values to luminance follow DICOM. * DICOM specified xchange and presentation but left implementation to vendors. * DICOM allowed image processing or standardization to occur on computer, graphics card or in the display itself. y2001 * DIN standard - mandatory in Germany * Assessed per 1) viewing conditions 2) grayscale reproduction, 3) spatial resolution, 4) contrast resolution, 5) line structure 6) color aspects, 7) artifacts, 8) Image instabilities * inclusive of acq. device and display device * Test images (including SMPTE) were specified * Signal via signal generator (analog systems) or Dx file. * 3 classes of devices cat A) Dx rad images cat B) all other images cat C) alphanumeric/graphic or control monitors *requirement for ratio of max to min luminance: cat A) 100 to 1 cat B) 40 to 1 *spatial luminescence (fractional dev between corner and center luminance) <30% CRT and within +/-15% for flat panel displays *IEC 61223-2-5:1994 luminence meaurements be made with a meter with absolute meas. uncertainty (2 rho) of 10%, range 0.05 cd/m2 to >=500 cd/m2, angular acceptance between 1 and 5 degrees, and photopic spectral sensitivity. ISO 9241 AND 13406 Series * Ergonomic Req. for Office Work with VDT Part 3 * Various standards for office class monitors, geometric linearity, orthogonality, minimum display luminance, minimum contrast, luminance ratios between hard and soft copy images, glare, luminance spatial uniformity, temporal instability (flicker), spatial instability (jitter) and screen image color. VESA Flat Panel Display (FPDM) Standard * helped establish a working FP metrology, later extended to CRT through FPDM version 2.0 2001 SECTION TWO - OVERVIEW OF ELECTRONIC DISPLAY TECHNOLOGY --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- * MI workstations have several physical and functional components * GP :Computer *CPU, coprocessor, i/o controllers, network comms * HID * Storage (HD)/DVD/CD * Output devices (printer, display, speakers) * Software; * Display controller (hardware, software) * OS * User application * Medical workstation *superior display interface, e.g. software, HR display, high performance display controllers * Operating System * Makes computer work and manages resources * * Display Processing software * Dx images = Dx grayscale array (e.g. CR, DR) or Dx data array (e.g. CT) * array to p-values to DDL done at either the OS level (embedded image processing) or at application level. * color is complex, usually done at apps level. *Display Controller (CRT-land): * display controller (video card, graphics card) is comb. of HW and SW to xform DDLs to appropriate signals for display dev. * Analog devices: * Has video memory that accepts output of the apps in screen-ready form. * Dx in video memory is xformed into signals ready for the display device. * Repeated scans of the video memory refresh the picture. * Computer has driver SW with interface to allow app to control contents of the video memory * Most current display devices accept only analog video signals (VESA 2001) * display controller performs a D-A conversion as the memory is scanned. * By driving the display device from DA converter, 2^n different voltages can be generated wherere n is the number of bits per pixel in the video memory. * Colour displays, 3 parallel DA convers for each pixel create RGB * # buts per pixel limit what is available to the display application and determines max num of shades of gray or colors * Video memory has 8 or more bits per pixel. * 8 bit grayscale controllers, 256 shades of gray are possible * RGB monitors, 8 bits/channel = 2^24 colors that can render 8-bit grayscale * Individual monitors respond differently to different voltages, so voltages are not evenly spaced * control of light output depends on being able to change the digital values * this control is necessary in MI displays * usually just for monochrome monitors (2010???) * MI displays: 10 - 12 bit memory and can store LUT to change the DDL in the memory for DA conversion * Use of proper LUT can make grayscale response follow a standard * Advanced controllers integrated luminance probes and calibration software to compute proper LUT * Consumer grade graphics cards (usually 8 bit memory): * Not suitable for most medical applications ` * LUT process may result in loss of distinct luminance levels, * 20 luminence steps are sacrified when correcting CRT and LCD monitors to DICOM GSDF function *Display Controllers (Flat panel) * DDL to luminance conversion changing (cite c2001) * Flat panel devices: * The controller sends Dx signal to the display device and the device converts this to signal to control luminesence * Basic requirement to standarize the relationship between DDL and luminance is the same. *Display Device * CRT vs Flat Panel *Photometric Quantities * Luminance * Rate at which visible light is emitted from a surface. * Energy of visible light emitted per second from a unit area on the surface into a unit solid angle (steradian). * Energy determined by standard weighting function * SI unit for energy of light is lumen-second, * Luminance: 1 lumen per steradian per m2 (candela per m2 (cd/m2 absolete = nit) * Lambertian distribution: * luminous intensity varies from a surface as the cosine of the viewing angle, * the appearance of surface brightness appears constant * example: matte painted white wall * Illuminance * rate at which visible light stikes a surface. * lumen/m2 or lux * Illuminence and luminence can be related to an ideal lambertian (reflective) surface * Illuminance of 1 lux striking a perfectly reflective surface will cause the emission of 1/pi observed luminance CRT Display (monochromatic or BW): * Filament: Heat source * Cathode: Electron source through thermionic emission * G1: control grid aperature (~1kV) * G2: accelerating electrode ~25kV * G3, G4, G5: lens system to bring electron beam to focus at phosphor * anode: aluminum coating starting near phosphor and back past deflection coils * emission is Lambertian (approx) * deflection coils (horizontal and vertical); control beam position on the phosphor * sawtooth generators sweep the beam horizontally and vertically at required frequencies * Phosphors: * tradeoffs: brightness vs resolution vs noise vs lifetime of phosphor effeciency vs persistance (screen lag or streaking) * P4 and P104 * maximum brightness 300 to 500 cd/m2 (mammo film viewing minimum 3000 cd/m2) * Modulation of voltage between G1 and cathode creates intensity variations CRT Display (colour): * Basically similar to BW, but there are 3 electron guns to make 3 scanning beams * Each beam aimed at its subpixel (RGB) * Design #1: guns are placed 120 degrees apart. * dot triad: Shadow mask with openings to permit each distinct RGB beam to pass to hit its subpixels * prone to veiling glare due to scattered electrons * Design #2: in line guns * Aperature grille vertically aligned slit aperatures. * also prone to veiling glare but not as bad as dot-triad * Combination of sub-pixels makes impression of color * veiling glare limits resolution of color displays, especially problematic when looking at grayscale images MI workstation CRT: * TV: 480 lines, 2 interlaced fields of 240 lines each to produce a frame every 1/30th of a second (30 fps). Paint time 53us * Workstation: 2000 lines, non-interlaced, 70 images per second. Paint time 5 us Video Signal: *Luminance created by * varing voltage between control aperature (G1) and cathode (K) * G1 is kept constant and K is driven (most common) * Brightness and contrast: * Contrast: Amplified video signal to emphasise luminance difference between pixels * Brightness: Controlled by a bias current to make an equal shift in luminance * In practice C and B are inter-related and must be tweaked together Pixel Characteristics * CRT images generated by scanning electron beam across phosphor. *sweep is continuous, so sharpness is dictated by pseudo Gaussian distribution * FLAT PANEL DISPLAY (FPD) matrix: * Pixels are descrete and one to one is preserved * In reality, each pixel is smaller than nominal because of electronic space * Ratio of active pixel area to nominal area is known as the aperture ratio (fill factor in FPD) * CRT pixel formation: * More complex tha FPD * Display controller does DA conversion, and coordinates display * scanning and intensity modulation achieved through 3 major signals: video, H-sync, V-sync * timing signals to suppress beam during line and field/frame retrace. * Smallest detail in CRT display limited by * 1) Hvideo(f): response function of display controller and cRT video circuits, * how fast the beam intensity can follow the voltage of the signal as it moves across a pixel * rise/fall time = time to change from 10% (almost black) to 90% (almost white). * 2) The nonlinearity of the relation between luminance and video signal voltage * 3) Motion of electron beam as affected by the beam deflection circuits, e.g. temporal input signal to spatial domain * 4) Hspot(f): response function of beam spot size formed by the electron optics on the phosphor screen e.g. parameters, beam current, phosphor layer thickness, scatter * Example: *display of an image 2048 x 2560 refresh rate of 71 Hz >Blanking and video signal delay = 26% time for frame (assumed) >To preserve detail, rise/fall time of pixel should be about 1/20 pixel time > That would need a bandwidth of about 2.5 GHz > Actual response time = 500 MHz peak (c 2001) > Result of BW limitation, pixels are larger than nominal. > Fortunately band-limited Dx images are less stringent, with BW limited by Nyquist * 5) Pixel size affected by beam current *related to emissive area on cathode * emissive area related to voltage diff between K and G1 * as EA increases current increases, beam diameter increases, spot increases > electrostatic repulsion of electrons in beam may bloom it but current dominent factor * As CRT ages, * phosphor ages, *beam current must be boosted to maintain luminance = loss of resolution * cathode depletes, * emissive area and more drive needed to maintain luminance = loss of res * Cathode del * 6) Pixel size affected by incident angle to the phosphor * resolution loss at edges of display because of deflective distortion (tear drop shaping) * Pixel astigmatism at display periphery * megapixel monitors (5 Mpx +) may use dynamic astigmatism compensation (return to round pixel) * 7) Colour CRT * shadow mask (dot triad) or aperture grille (bars) add to limits in spatial resolution * act as a sampling comb, phosphor layer is not continuous ` * The electron beam covers 5 - 10 sets of pixel islands (RGB subpixels). * Spatial resolution of color CRT is much poorer than grayscale * Alternate tech to CRT * AMLCD (currrent 2010), transmissive tech * OLED (heir apparent 2010), emissive tech * FED and OFED Field-Emitter Display (upcoming with organic tech??) emissive tech? * micromirrow displays (holographic?) * plasma displays (good tech but apparently on life support 2010) * electronic projection displays (?) and head-mounted displays (?) * LCD * Molecular orientation of LC altered by external electric field - changes optical qualities and modifies light transmission * Backlight + polarizer filter + large array of LC cells + exit polarizing filter * height and volume of LC cells controlled by spacers * Tx intensity relates: > Backlight intensity, > 1st polarizing filter, applied twist in LC, 2nd polarising filter, aperture ratio (transparent part of LC) > For color: color filters transmissive properties * LC controlling voltages through TFT array * LCD characterized seen as normally white and normally black - depends on LC natural "twist" * Eg 1: If crossed polarizing filters + LC no intrinsic twist+ no voltage = light is blocked and display is black * Eg 2: If colinear polarizing filters + LC no intrinsic twist + no voltage = light Xmits and display is white * White display = more luminence, but difficult to get a decent black (Lmin) because crystal twist has to be to max. * Light emission is non-Lambertian. * optical anisotropy of the LC cell (rel. to design, and Vapp) * use of polarizing filters * sine squared effect - potentially severe angular dependence of luminence with contrast loss as well * sine squared effect minimized by; >1) varying molecular alignments in subregions (domains) within pixels >2) modifying the orientation of the LC molecules to remain in the plane of the display (inplane switch) >3) adding a negative birefringence plan to compensate for anisotropy > Usually with 1 and 2, 3 is used * Fielld Emissive Displays (FED): * mini CRTs using cold source emission instead of thermionic emission, 1 emitter + vacuum cell + phosphor/cathode * Favorable characteristics: > Good temp and humidity tolerence, wide viewing angle with Lambertian emission (like CRT), > Bad: severe pixel luminance non-uniformity - hard to standardize cathode emission, low reliability * Organic Light-Emitting Displays (OLED) *Electroluminescence (EL) all solid state approach providing most direct conversion of electrical energy into light. * OLED better that classic LED * Superior light emission, difficulty producing white (2001, also 2010) *Engineering specs for Display Devices TYPICAL SPECS - PERFORMANCE REQUIREMENTS ARE LATER * Primary Monitor: > Matrix size: 1728 x 2304 (3 - 4 Megapixel) > Active pixel size, mm 0.17 - 0.23 > Luminance ratios: 250 > Luminance non-uniformity: <25% > For CRTs: >Amp BW 250 - 290 MHz at 45V p-p > Phosphor type P45 >Max Lum cd/m2 200 - 300 > RAR (specified px format) 0.9 to 1.1 > Pixel size at 5% pt <2.5:1 to 50% > For LCDs >Max Lum cd/m2 700 > Viewing angle (@40:1 lum ratio) >80 deg. hor, 50 deg vert > Defective pixels <10 * Secondary/Primary Monitor: > Matrix size: 1200 x 1600 (1 -2 Megapixel) > Active pixel size, mm 0.28 - 0.3 > Luminance ratios: ~ 100 - 250 > Luminance non-uniformity: <30% > For CRTs: >Amp BW 160 - 200 MHz at 45V p-p > Phosphor type P104 or P45 >Max Lum cd/m2 100 - 300 > RAR (specified px format) 0.9 to 1.1 > Pixel size at 5% pt <3:1 to 50% > For LCDs >Max Lum cd/m2 200 > Viewing angle (@40:1 lum ratio) >80 deg. hor, 50 deg vert > Defective pixels <10 1) Matrix size = addressable pixels not what it will resolve 2) Luminenance ratios = end points (factory settings) black black and white white DAC of 0 and 255 8 bit 3) Luminenance non-uniformity = Dependent on glass formula and bulb type. Compensation for this may drop resolution at the ends of the display due to increase beam current and blooming 4) Anti-reflective coatins reduce specular reflectance and veiling glare in CRTs. Multilayer is best 5) Misc specs, e.g. non-linearity <=10%, 0.05mm max, 6) Amp BW is alternately known as 3 db at volts p-p 7) P45 best phosphor for stability, low spatiial noise 8) Max luminesence: specified at a specific pixel format 9) RAR = Resolution-Addressability-Ratio. Measured pixel at 50% * Luminence Uniformity > CRT: many factors - glass thickness, type, angle of incidence on phosphor > LCD: backlight non-uniformity from light source, variations in thickness of LC layer, no resolution dependance like CRT * Surface treatments > LCD, spacers that are light absorbers (like grids in x-ray), antiglare faceplates but these can cause haziness * Bit depth * Note: most diagnostc monitors are 12 bit now * Theoretical shades of gray vs actual may vary * insufficient BW masks tonal values (CRT) * LUTs can significantly reduce the actual number of luminence values commanded by an 8-bit display controller * Viewing angle (FPD): * LCD have an angular dependent luminance and contrast reponse and chromaticity, even within its specified viewing angle * Aperture Ratio (LCD) * Actual pixel size vs nominal pixel size. High ratio = more pixel, less filler * Classification of Display devices * Primary or diagnostic: used to interpret images, e.g. radiology, ortho * Secondary or clinical: used for other than interpretation or diagnosis