technology meets pathology Institute of Pathology, Charité Universitätsmedizin Berlin, Berlin, Germany 3 Overview

Transcription

1 ASSESSMENT OF TECHNICAL PARAMETERS A. Alekseychuk 1, N. Zerbe 2, Y. Yagi 3 1 Computer Vision and Remote Sensing, TU Berlin, Berlin, Germany 2 Institute of Pathology, Charité Universitätsmedizin Berlin, Berlin, Germany 3 Massachusetts General Hospital, Boston, United States Overview Non-medical parameters evaluated Precision: technical image quality Colour fidelity & resolution Geometrical distortions and effective pixel size Green IT: environmental/ergonomic properties Noise emission Power consumption 2

2 Power consumption Motivation: Large range of illuminators - Incandescent (halogen) bulbs: W - High intensity gas discharge lamps (xenon, mercury, tin-halide): W - LED: 3-5 W typical Mechanics and electronics - slide advance & autoloader - CCD camera (10-15W non-cooled) Measuring: - in-socket device - ±1%, ±1W typical 250 W 5 W 3 Power consumption measurements 40x scan ( Quality ) of about 1 cm² specimen Scanner only (without computer and monitor) Results: Power range: 19 W 360 W Energy per scan: 6 Wh 83 Wh Highest Wattage not highest consumption 4

3 Noise emission measurements Acoustic Camera (gfai tech GmbH, phased array of 48 microphones fine spectral and time resolution (down to 10 µs) distribution of sound emission sound image simple video camera for overlay 5 Noise emission measurements Measured during Quality scan Absolute measurements difficult (background noise) 6

4 Geometrical precision Aims of investigation: determination of true effective pixel size image distortions no stitching assessment Main test specimen: overall size: 1 x 3 (25mm x 75mm) image area: 20mm x 50mm clear aperture: 8.5 µm² opaque lines: 1.5 µm² pitch: 10 µm 7 Geometrical precision Other test specimens

5 Colour and its measurement Human: trichromatic vision Cones spectral sensitivity nor RGB neither XYZ, etc. complex mechanism Digital sensing accuracy spectral properties of sensor encoding and calibration Sufficient illuminator CRI / CIE metamerism index Danger: slight shades of colour not distinguishable 9 Spectra of light sources Tungsten lamps (halogen): smooth spectrum, 3200K typical Arc lamps (xenon): flat continuous spectrum, 6500K (daylight) Fluorescent lamps (LID): only a few spectral lines (usually) problems LED: fundamentally monochromatic (single mode) with luminophor wider 10

6 Colour accuracy test method Aims of assessment : - colour fidelity of devices absolute accuracy of colour reproduction colour resolution detectability of subtle differences IT8.7/1 colour target colour & 24 grey value fields known absorption spectra known colourimetric coordinates Testing method 1. perform low-resolution scan 2. average inside middle 50% of each field 11 Colour accuracy test method 3. Chromatic adaptation (property of human vision) colour temperature correction after Von Kries 4. Colour difference calculation to CIEDE2000 definition (resolves CIE94 de perceptual uniformity issue) contours of justplanckian locus and constant CCT lines noticeabledifferences (JND) 12

7 Colour accuracy test method Fidelity test: average de over 144 fields de avg =n de(c * mes, c ref ) /144 the lower the better mix-colours matrix for fidelity test Colour resolution ability: minimum over step differences on step wedges de step = min de(c mes (i), c mes (i+1)), i [B,J] the higher the better colour step wedges Conclusions 1st time technical domain 5 of 6 manufactures participated Performed a few simple yet useful tests First time metrological assessment of colour fidelity of histological scanners technical limits of visibility of structures Outlook: more quantitative methods more metrology in this domain (60dBA vs. 65dBA 60 ±6dBA vs..) IT8 histological dies, matrix of application-specific colours advanced geometry (e.g. evaluation of quality of stitching) 14

8 Thank you for your attention! 15 16