Some EBT2 problems solved

Comments Off on Some EBT2 problems solved

Observed problem 1:

Interference artefacts become apparent when the film is placed on the glass scanning bed (see Figure 1). (These “Newton’s rings” can be seen without scanning the film, can be moved around when the film is moved or touched, and are also apparent when two sheets of film are placed on top of one another.) “Calculating dose as the average across a region of the image that is subject to Newton’s rings can lead to an dose uncertainty of ±5% (plus or minus one standard deviation), in regions of high dose, and can lead to a 30% increase in the dose measured in out-of-field regions” [1]

Figure 1: Red-channel images of an un-irradiated sheet of EBT2 film, enlarged to show scanning artefact.

Suggested solution 1:

All scans are performed with the film resting on top of a 1.5 mm thick plastic frame, to keep them out of contact with the glass surface of the scanner.

Observed problem 2:

Banding effects appear in “early batches” of EBT2 film. Some users report that banding can be observed by simply looking at a sheet of film, fresh from its packet. For affected films, the banding is certainly apparent once the film is scanned, if the window levels are adjusted appropriately (see Figure 2). The heterogeneity of the film can lead to a delivered dose of 2 Gy being measured as 2.1 Gy in a region of high optical density or 1.9 Gy in a region of low optical density. This amounts to a measured dose difference of 10% between light and dark bands.

Figure 2: Red-channel images of an un-irradiated sheet of EBT2 film (at four different scanning orientations), with window levels ajusted to show film heterogeneities.(Each image is approximately 20 x 20 cm2.)

Suggested solution 2:

All films are scanned both before and after irradiation, with the optical density calculated using the ratio of the pre- and post- irradiation images, at each pixel:

netODred = log [ Pred,pre(i, j) / Pred,post,C(i, j) ]

where Pred,pre is the pixel value in the pre-scan red-channel image and Pred,post,C is the pixel value in the scanner-output-corrected post-scan red-channel image.

Figure 3: Profiles across an out-of-field region: (left figure) pre- and post- irradiation pixel values and (right figure) the ratio of pre-irradiation pixel values to post-irradiation pixel values. (The horizontal scale in both figures is centimetres.)

The use of this pixelwise evaluation of net optical density means that any constant spatial variation in scanner output is intrinsically corrected. An additional correction for overall scanner output variation from one scan to the next can be calculated using the ratio of the mean pixel value in a region of interest in an unexposed (UE) piece of film, before and after the rest of the film is irradiated, to give

Pred,post,C = Pred,post x (Pred,pre,UE / Pred,post,UE).

Apart from correcting for scanner output, this method also corrects for the centimetre-scale oscillations in pixel value that result from the heterogeneity of the film itself. This can be seen in Figure 3, where the illustration on the left shows pixel value oscillations over a much larger scale than the apparent noise, and where the illustration on the right shows that these oscillations have been corrected. (Note that these profiles are for an out-of-field region where the SNR is very low.)

The use of this technique to correct for film heterogeneity assumes

(a)  that the effects of the film heterogeneity do not vary with delivered dose, and

(b)  that the alignment on the scanning bed of the film before and after irradiation is perfectly reproducible.

The following is a preliminary indication of the veracity of assumption (a): Our calibration films, which were irradiated with doses ranging from 0 to 400 cGy, and which were scanned and evaluated according to our proposed procedure, are all more-or-less equally unaffected by banding.

By contrast, assumption (b) has been found to be invalid (if slightly). Registration of the written marks (labels and numbers) that we made on the film indicates that our post-irradiation scans are can be displaced by up to 0.8mm relative to our post-irradiation scans. Our system of aligning to marks on the plastic frame is simply not perfectly reliable. However, since the heterogeneity of the film appears as centimetre-scale bands, a submilimetre displacement in the film scanning positions is regarded as insignificant. (It’s effect, at least, is not apparent in the resulting dose maps.)

Proposed EBT2 film scanning method:

The following quotation describing our scanning and analysis method is taken from Ref [1] (with the exception of the photos; the photos are new):

While avoiding the blue-channel correction of the red-channel optical density, the appearance of spatial variation in the measured dose that would otherwise result from heterogeneities in the film itself can be reduced by using the following procedure:

For each piece of film to be irradiated:

(i)            Cut two smaller pieces from the film to be used as references. (For example, each 20.4× 25.4 cm2 sheet can be cut into one 20.4 × 20.4 cm2 piece and two 10.2× 5.0 cm2 pieces of film.)

(ii)          Pre-scan all three pieces of film, together, prior to irradiation.

(iii)         Irradiate the larger, ‘measurement’, piece of film with the dose to be measured and irradiate one of the smaller, ‘reference’, pieces of film with a known dose. Leave the other reference piece of film un-irradiated.

(iv)         Post-scan all three pieces of film, together, after waiting for the irradiated films to develop (at least 6 hours but preferably 24 hours after irradiation).

(v)          Obtain red channel images from the pre-scan and post-scan images.

(vi)         Use mean pixel values from a region in the centre of the un-irradiated reference film, from the pre-scan and the post-scan red-channel images, to obtain a correction for any scanner output variations that may be caused by variations in the scanner light source intensity.

(vii)        Convert value in each pixel (i, j) to net optical density, using the relationship

netODred = log [ Pred,pre(i, j) / Pred,post,C(i, j) ] (1)

where Pred,pre is the pixel value in the pre-scan red-channel image and Pred,post,C is the pixel value in the scanner-output-corrected post-scan red-channel image.

An improved method for irradiating and scanning the calibration films is currently under development.


1. T. Kairn, T. Aland, J. Kenny, Phys. Med. Biol. 55(15): L37-L42 (2010)