GB2397959A

GB2397959A - Imaging system using software model to correct for thermal dark current noise.

Info

Publication number: GB2397959A
Application number: GB0225819A
Authority: GB
Inventors: Kevin Wells; Eleni Kokkinou
Original assignee: University of Surrey
Current assignee: University of Surrey
Priority date: 2002-11-06
Filing date: 2002-11-06
Publication date: 2004-08-04
Anticipated expiration: 2022-11-06
Also published as: GB2397959B; GB0225819D0

Abstract

In an image system such as that used in autoradiography, a software model or algorithm for inter pixel noise and interframe variations in operating bias is used to correct for thermal dark current noise. As shown, the autoradiography imaging system comprises an image sensor 1, such as a CCD to capture low radiation levels, housed in a light proof enclosure 2, associated readout/control electronics and a computer 4, running software to reduce noise. By using a mathematical software method to remove pixel offsets caused by dark current, operation without sensor cooling is allowed.

Description

1 2397959

IMAGING SYSTEM FOR USE IN AUTORADIOGRAPHY

1. Background

The invention relates to an imaging system that consists of an imaging sensor such as a CCD housed in a light-tight enclosure, associated readout electronics, a computer to control the system and store the image data, and software to control the system's operation and to process the raw image data. The principal application is for imaging very low levels of radiation, as produced by radio- labelled tissue samples in autoradiography. Other applications include imaging of low light levels and radiation imaging and detection as might be found in the nuclear industry and biomedical research.

Autoradiography generally refers to the imaging of radioactive specimens such as thin slices of tissue, usually undertaken by holding the specimen in intimate contact with X-ray film or film emulsion. However, the low levels of radioactivity involved, combined with the relatively low sensitivity of film to detecting this radiation means that exposure times can last for up to several months, although a few days is more typical. To address this and other disadvantages associated with using film, a number of other imaging systems have been developed which can produce autoradiographic images in a fraction of the time compared to undertaking imaging using conventional X-ray film. However, these tend to be expensive because the technology involved is complicated. By contrast the system described here uses low cost, widely available image sensor technology, such as CCDs, and uses a mathematical correction in order to correct the noise corrupted raw data generated by the imaging device.

2. The Problem Addressed by the Invention The conventional method of correcting for thermal noise in many imaging devices is to cool the actual device. This reduces thermally generated noise (dark current) to minimal or negligible levels. It was previously thought that without this step, using an imaging device such as a CCD at room temperature with long exposure times would be impossible for low level imaging applications (for example imaging tissue that is weakly radioactive). This is because the thermal dark current noise generated at room temperature will be of similar magnitude to the desired signal, particularly where long exposure times are used.

Imaging applications which generate very low levels of signal would then be effectively swamped by the noise, thus obscuring any useful image information.

However, by modeling the individual pixel response and also by correcting for inter-frame variations in operating bias, then noise corruption due to thermal dark current and potentially other sources of noise, can be effectively minimized.

This allows useful images to be produced by the system at normal room temperature in full frame- slow-scan -mode, or normal video rates, using standard sensor technology, and without further recourse to any specialist developments in sensor technology.

3. Description of The Invention

A preferred embodiment of the imaging system appears in Figure 1. The imaging sensor[1] is housed in a light proof compartment[2] with electrical connections leading to the control electronics [3]. Within the housing is a first level of amplification and buffering before the sensor signal data is passed to the control electronics. In this particular case the control electronics includes electronic hardware which allows the user to set the clock speed used for readout and the exposure time per frame. It also includes proprietary noise suppression hardware (e.g. double correlated sampling) and several levels of amplification and buffering before the raw data, in the form of analogue pixel signals, are passed to the computer [4]. This is used to control the system and store the image data. The computer utilizes a mathematical correction which is applied to the raw image data to remove pixel offsets caused by cumulative dark current.

This is described in section 4.

a À Àe À.

À e . . .. À .. À

Operation of the imaging sensor is controlled by the control electronics [3] with settings for clock speed and frame exposure time being set by the user. A preferred setting would be a lMHz clock speed and 15 seconds exposure time per image frame. To undertake an imaging experiment, first a sample is placed directly on the image sensor's surface [1], and the compartment [2] is sealed.

Once image acquisition commences, image frame data are transmitted to the computer [4] via a suitable connection and processed using the correction described below. Individual image frames are then binarised and summed to produce a composite image. Where necessary the summed image may be windowed or thresholded to better visualize the image data.

4. Detailed Description of the Correction Method Employed In order to correct for thermal noise, a priori knowledge of the individual pixel's behaviour must be known. Therefore, a set of reference blank frames is thus required, before imaging of a radioactive or other type of radiation source can commence. Typically, 500 blank frames of data are required to ensure consistent results. From these blank frame data, the modal value for each pixel, ry, is calculated, along with the average mode value, R of the entire data set over the 500 frames.

The correction is based on the assumption that only a low level radiation source is being imaged. This means that most pixels contain only thermal noise rather than signal generated by the radiation source. A further assumption used in the correction is that thermal noise is the dominant source of noise in the system.

Therefore, the peak in the intensity distribution of pixels across the frame corresponds to a peak generated by thermal dark current noise.

Once a frame of subsequent raw image data are available, then any individual pixel, at location in, in the frame can be corrected by using a predicted value p À . À À. . e À . . . . . À . À . . . . of the dark current noise component obtained from a linear relationship, of the form pjj=mrij +c, where m and c are constants, and rv represents the mode value obtained from a reference blank frame data set, or obtained as a linear function of the mode of the reference frame, R. A preferred parameterization of this relationship is shown equation 2 below. The predicted value pit for a pixel dark current offset at location id is thus obtained from: k r + G - R. . .2

_

where G represents the modal value of subsequent experimental frame, R represents the average modal value from the blank frame data and rv represents the mode of an individual pixel at location if from the reference data set over the 500 frames. The term kv represents a gain term which accounts for each pixel's individual response to changes in ambient temperature.

The raw data can thus be corrected by simple subtraction leaving a corrected value cv: CV 9'j - Pin. . . 3 where go represents the raw uncorrected recorded intensity of the pixel at location if from the raw uncorrected experimental data.

The mode is used, rather than the mean, as this is a more robust estimator of relative thermal shift in operating point. This is because use of the mean may shift the estimate of the peak in the noise distribution due to the weighting caused by high intensity broken pixels or due to pixels containing signal charge.

À . À . À À À ÀÀ* À ÀÀ ÀÀÀ : : : The system is therefore able to correct for any inter-frame shift in thermal operating point, which governs thermal noise generation, and also corrects for the relative offset of each pixel in the image by reference to a blank frame data set. This processing of the data can reduce pixel dark current by up to 55%.

À . À. À.

À À À À . . ... ... ..

À À À . À À . À À e À. e

Claims

C L A I M S

1. An imaging system, principally for use in autoradiography, which can operate without the need for cooling by using a software model for inter pixel noise and inter-frame variations in operating bias to correct for the effects of thermal dark current.

2. A system, as claimed in Claim 1, which can image very low levels of radiation (down to a few kBq) with linear response over at least 2 orders of magnitude, whilst tolerating wide variations in pixel dark current generation and thermal shifts in operating point of several 10s of degrees.

À . À .. À.

À À . À À À À . . .