• Decrease font size
  • Return font size to normal
  • Increase font size
U.S. Department of Health and Human Services

Radiation-Emitting Products

  • Print
  • Share
  • E-mail

Dose and Image Quality in Mammography: Trends during the First Decade of MQSA

(Related Article: Updated Trends in Mammography Dose and Image Quality)

 

David C. Spelic, Ph.D.
US Food and Drug Administration
Division of Mammography Quality and Radiation Programs
1350 Piccard Drive HFZ-240
Rockville, MD 20850
Phone: 301-594-3525
Email: dcs@cdrh.fda.gov
9/5/2003

 

Introduction

 

The US Food and Drug Administration (FDA) assures the quality of mammography conducted in the United States through annual Mammography Quality Standards Act (MQSA) inspections of facilities performing this screening x-ray examination. While there are many facets to the overall practice of mammography, improvements in the technical features of this exam over the past several decades have resulted in better quality from a radiological health standpoint. One of the primary concerns of mammography during its infancy was the need to administer relatively high doses to achieve image quality that was at best marginal. The mammography community realized that the benefit relative to the associated risk had to be improved. MQSA inspection results, along with the Nationwide Evaluation of X-ray Trends (NEXT) mammography surveys, have documented trends in technical improvements on both sides of the equation: risk (radiation dose) has decreased with time while the associated benefit (the quality of the resulting radiograph) has improved. This article highlights recent trends in selected technical indicators of quality in mammography.

The FDA has previously reported the results of NEXT surveys on technical features of mammography that occurred prior to the beginning of MQSA inspections, and for MQSA inspections conducted through 1997. In their discussions of NEXT mammography surveys conducted in 1985, 1988, and 19921 , Conway et al. documented a trend toward higher dose for screen-film systems, but they associated this with an increased use of grids, leading to substantially improved image quality. In a second paper discussing the results of early MQSA inspections2 , Suleiman et al. observed further improvements in technical areas such as phantom image quality scores, reduced darkroom fog, and improved film processing quality while also noting that doses were rising slightly. Mammography is likely one of the most difficult of diagnostic radiographic exams to administer from a technical standpoint because of its demand for both high spatial resolution and high film contrast while keeping the administered dose acceptable. These requirements challenge the image quality characteristics of the imaging system, including the type of film and screen selected, the film’s operating point (ie. its background optical density), and the quality of the film processing.

Since the beginning of MQSA inspections in 1995, there have been noticeable changes in the technical aspects of mammography. Table 1 lists trends in selected parameters during the course of the first nine years of inspections.

1995 1997 1999 2001 2003*
Dose (mGy)**
1.50 1.60 1.65 1.76 1.76
Phantom Score
11.5 11.7 12.0 12.2 12.2
Processing Speed (std)
97 103 101 104 111
Phantom Image Background
Optical Density
1.42 1.52 1.63 1.70 1.78
Darkroom Fog
0.04 0.03 0.03 0.03 0.02

*Data for calendar year 2003 is for inspections through the end of April.
** To convert the unit of dose to mrad, multiply by 100 mrad/mGy.

Table 1. Mean values for selected technical parameters observed during MQSA inspections.

The table above shows some interesting trends. First, we observe a modest improvement in the average phantom image score. We also note that the average phantom image background optical density has increased quite significantly. Similarly the indicator for film processing quality, the processing speed, has also increased. Finally, there has also been a small increase in the average mean glandular dose. Can we draw any hard conclusions from these statistical parameters? For instance, is the increase in mean phantom score due primarily to significantly fewer facilities with low scores, or is it due to general improvement by the practicing community? Are such changes significant? We could pose a similar question for the mean glandular dose. To answer these, and similar questions, we must look at the distributions of these mammography parameters.

 

Mean Glandular Dose

 Figure 1 (table A-1) shows the distribution of mean glandular doses for three periods of time between 1995 and 2002. Inspection data from 1995 and 1997 are displayed to illustrate the distribution of dose during the first inspection, and after at least one inspection respectively. Data for 2002, the most recent complete calendar year of inspections, is also plotted. Two observations can be drawn from these distributions. The distribution of mean glandular doses shifts toward slightly higher values with time. Additionally, the spread of the dose values about the respective mean has narrowed. Not only are facilities administering a moderately higher dose on average, but the dispersion of doses across all facilities has decreased with time. Statistical testing in fact shows that the difference between the two distributions for mean glandular dose for the time period from 1995 to 2002 is statistically significant (p < 0.001). In other words the difference between these two distributions is not a coincidence, but rather derives from the fact that facilities inspected in 2002 were indeed administering higher doses on average than the facilities inspected in 1995. Why would there be a trend toward higher doses? What is the benefit from higher dose? To answer these questions we will examine several technical aspects of mammography that relate to the dose administered to the mammography patient: 1) film processing quality, 2) the desired optical density of the film, and 3) image quality. We will first examine the quality of film processing.

Figure 1. Distributions of mean glandular dose from MQSA inspections

Figure 1. Distributions of mean glandular dose from MQSA inspections

 

Film Processing Quality 

The quality of film processing can directly impact on both image quality and patient dose. Earlier studies have documented the extent to which facilities conducting various types of diagnostic x-ray exams failed to meet minimum levels of film processing quality3. During MQSA inspections, the quality of film processing is evaluated using the Sensitometric Technique for the Evaluation of Processing (STEP) procedure. A film processor is assigned a numerical speed value in conjunction with the level of processing quality as compared with what the quality of processing would be if it were operating according to the film manufacturer’s recommendations. An ideal processor operating exactly as specified by the film manufacturer is assigned a speed of 100. Processors found to be under-developing their films result in speed values below 100, with a speed of 80 being the minimum level of acceptability. Built in to the 20% tolerance range is the acknowledgement that there are a variety of brands of mammography film in clinical use, whereas MQSA inspectors use a single brand of film for their testing.

 Figure 2 (table A-2) shows trends in film processing speed, and a subtle bimodal aspect can be seen for the two earlier distributions for 1995 and 1997. Facilities with processing speeds in the range of 120 and above were predominantly performing extended cycle processing. In this particular processing cycle, the film typically travels the processor at a reduced speed hence spending more time in the chemical environment. This results in a more complete development of the emulsion, and hence a darker film. At the onset of MQSA inspections, extended cycle processing was fairly popular: nearly 50% of facilities visited during the first round of inspections were performing extended cycle processing. Extended cycle processing has since become significantly less frequent, with fewer than 5% of facilities performing extended cycle processing in 2002. A likely reason for the significant decrease in extended processing over time was the discontinuation of a film type produced specifically for extended cycle processing by a major manufacturer, accounting for the film used in over 50% of all facilities performing extended cycle processing between 1995 and 1996. Facilities are encouraged to follow the film manufacturer’s recommendations, and the majority of manufacturers now specify standard cycle processing for their mammography films.

Figure 2. Distributions of film processing speeds from MQSA inspections

Figure 2.  Distributions of film processing speeds from MQSA inspections.

 

So how does the quality of film processing impact on dose? Processing speeds that fall below 80 indicate that the film processor is most likely not fully developing the emulsion, and is producing films of significantly lower optical density than would normally result if the film processor were operating close to a speed of 100. One could compensate for this by increasing the exposure to the film, but would thereby increase the dose to the patient (excluding a change in the screen-film combination speed). What should be expected in practice, however, is a decrease in dose over time as such facilities found to have low processing address the problem and find that their radiographic technique can be reduced.

If a facility converts from extended cycle processing to standard cycle processing while using the same film, they most likely would need to increase their radiographic technique in order to maintain a consistent film optical density. In fact, analysis of inspection data for 1995 shows that facilities using extended-cycle processing had significantly lower doses (p < .001) than facilities using standard-cycle processing. Hence we could anticipate an increase in dose associated with fewer facilities conducting extended cycle processing. Figure 2 also shows that over time, fewer facilities had film processors with a speed below 100. One could therefore argue that such facilities have improved their film processing quality and can reduce their patient dose. However this is a difficult conclusion to support because the processing evaluation considers any processing speed of 80 or above to be acceptable, and a moderate increase in processing speed can occur when a facility switches to a different brand of chemistry or to a different brand of film. If the processor is determined to be operating acceptably, then the facility may not be motivated to adjust the processor if other technical measures of performance (e.g. dose, phantom score) are acceptable.

Image Quality: Phantom Film Background Optical Density

 A study of the background optical density (OD) of the phantom film obtained during the MQSA inspection also reveals a trend that can contribute to the higher observed doses. Figure 3 (table A-3) shows a dramatic trend toward higher background film optical densities, with the distribution for 2002 showing very few facilities with background OD’s at or below 1.4 (selected as a value close to, but above the MQSA minimum OD requirement of 1.2 for the background region on the phantom film). The distributions of 1995 and 1997, however, show that a significant number of inspected facilities had background film OD’s at or below 1.4. To understand the impact of the background OD on the quality of the mammogram, one should keep in mind the sensitometric properties of radiographic film. Mammography film, much like all diagnostic x-ray film types, has a preferred range of exposure for which a diagnostically useful image will result. If the exposure to the film is either too high or too low, one loses image quality because the film is not responding to changes in exposure (subject contrast) with significant changes in film optical density (film contrast).

Figure 3. Distributions of phantom film background optical density from MQSA inspections

Figure 3. Distributions of phantom film background optical density from MQSA inspections

 

This behavior can be observed on films produced during routine film processing QC- the sensitometric strip will show almost no change in OD for the first few and last few steps because these exposures from the sensitometer are not within the film’s range of response. But is higher OD better, especially given that one needs to increase the dose to the patient (other relevant factors kept constant)? Most commercially available mammography films provide better film contrast at OD’s above 1.4, with some films performing well at OD’s approaching 2.04. KC Young et al. found in their investigation that one benefit of higher OD’s is an increased detection rate for small breast cancers5 .

Phantom Image Quality Score

 

 Given the observed shift toward higher OD’s, we should expect to observe a trend toward better image quality performance. Figure 4 (table A-4) shows a statistically significant improvement (p<.001) in phantom image quality scores between 1995 and 2002. Not only has the mean phantom score increased with time, but the distributions show that facilities are producing better films across the board. If we pull our observations together, we find that although the average for mean glandular dose is recently rising moderately- and this can be partly attributed to significantly fewer facilities using extended-cycle film processing- facilities on average are also producing films with higher background optical densities, which can result in better image quality. It is difficult to determine whether the increased benefit to the population associated with higher image quality scores is commensurate with the increase in dose because the phantom image quality score is merely one indicator of the overall image quality of the mammography exam.

Figure 4. Distributions of phantom film image quality score from MQSA inspections

Figure 4. Distributions of phantom film image quality score from MQSA inspections

 

Other factors must be considered, and ultimately the radiologist should play a significant role in determining the extent of increased benefit. However one can minimally argue that there is good potential for increased population benefit just from the fact that fewer facilities have phantom image quality scores near the minimum acceptable score of 10.0.

Summary

 In this article, we have discussed trends in technical features of mammography since the beginning of MQSA inspections. We have observed significant reduction in extended cycle processing, trends toward higher mean glandular doses and higher background optical densities, and improvement in phantom image quality scores. If we turn the clock back further however, and consider the changes mammography has undergone in the past several decades, we see even more dramatic changes to the practice. Figure 5 (table A-5) illustrates the trends in dose and image quality in mammography with time, showing how significant the changes in dose and image quality have been from the pre-70’s practice using industrial film with resulting high doses, to the era of dedicated mammography x-ray equipment and screen-film combinations that permit lower doses while providing improved image quality. In this context, the modest increase in dose observed recently may not be very significant. The technical aspects of screen-film mammography appear to be fairly optimized at present; however digital-based mammography technologies including computer-aided diagnosis (CAD) may offer further improvements in this important tool for the early detection of breast cancer.

Figure 5. Trends in mammography dose and image quality. Updated reprint of figure 1, reference 2. Note that the reported values for phantom image quality score are without artifact subtraction.

Figure 5. Trends in mammography dose and image quality. Updated reprint of figure 1, reference 2. Note that the reported values for phantom image quality score are without artifact subtraction. This permitted comparison between MQSA scores, and scores reported from NEXT surveys conducted in 1985, 1988, and 1992 where artifact subtraction was not reported. The 1985 NEXT survey scores are reported for the Radiation Measurements Inc. (RMI) model 152 phantom with the 'C' insert, equivalent to an approximately 4.7cm compressed breast. The 1988 survey scores are reported for the RMI 156 phantom with the 'C' insert, and the 1992 survey scores are reported for the RMI model 156 phantom with the 'D' insert.



Appendix: Tabulated data for figures 1-5.

 

 Dose (mGy)

1995

1997

2002

< 0.80

2

1

0

0.80-0.99

7

4

0

1.00-1.19

15

10

2

1.20-1.39

20

17

8

1.40-1.59

20

22

19

1.60-1.79

16

19

27

1.80-1.99

9

14

22

2.00-2.19

5

7

12

2.20-2.39

3

3

6

2.40-2.59

2

2

2

2.60 +

1

1

1

Table A-1.  Tabulated data for figure 1, mean glandular dose distributions. (back to text.)


 

 Speed

1995

1997

2002

Below 80

3

1

0

80-89

14

8

3

90-99

24

23

20

100-109

17

22

41

110-119

10

15

24

120-129

10

13

9

130-139

9

9

2

140-149

7

4

1

150-159

4

2

0

160 and above

2

2

0

Table A-2. Tabulated data for figure 2, film processing speed distributions. (back to text.)


 

 Optical Density

1995

1997

2002

Below 1.20

10

4

0

1.20-1.39

33

21

2

1.40-1.59

37

40

15

1.60-1.79

15

25

40

1.80-1.99

3

7

32

2.00-2.19

1

1

8

2.20 and above

0

0

1

Table A-3. Tabulated data for figure 3, mammography inspection phantom film background optical density distributions. (back to text.)


 

 Phantom Score

1995

1997

2002

 9.0 and below

1

1

0

9.5-10.0

2

2

0

10.5-11.0

19

15

6

11.5-12.0

37

33

23

12.5-13.0

29

32

36

13.5-14.0

10

15

27

14.5 and above

1

3

7

Table A-4. Tabulated data for figure 4, phantom image score distributions. (back to text.)


 

 Year

Dose

IQ

1974

13.8

-

1976

4.2

-

1980

2.5

-

1985

2.2

7.8

1988

1.8

10.3

1992

1.5

11.2

1995

1.5

11.9

1996

1.6

12

1997

1.6

12.2

1998

1.6

12.3

1999

1.7

12.5

2000

1.7

12.6

2001

1.8

12.7

2002

1.8

12.8

Table A-5. Tabulated data for figure 5, mean glandular dose and image quality trends by year. (back to text.)


 

 

References

 


  1. Conway BJ, Suleiman OH, Rueter FG, Antonsen RG, Slayton RJ. National Survey of Mammography Facilities in 1985, 1988, and 1992. Radiology 1994; 191: 323-330.
  2. Suleiman,OH, Spelic DC, McCrohan JL, Symonds GR, Houn F. Mammography in the 1990s: The United States and Canada. Radiology 1999; 210: 345-351.
  3. Suleiman OH, Conway BJ, Rueter FG, Slayton RJ. Automatic Film Processing: Analysis of 9 Years of Observations. Radiology 1992; 185: 25-28.
  4. West MS, Spelic DC. Using Light Sensitometry to Evaluate Mammography Film Performance. Med. Phys. 27(5), May 2000: 854-860.
  5. Young KC, Wallis MG, Ramsdale ML. Mammography Film Density and Detection of Small Breast Cancers. Clinical Radiology (49), 1994: 461-465.