ORIGINAL RESEARCH • BREAST IMAGING

Cancer Yield and Patterns of Follow-up for BI-RADS

Category 3 after Screening Mammography Recall in the
National Mammography Database
Wendie A. Berg, MD, PhD • Jeremy M. Berg, PhD • Edward A. Sickles, MD • Elizabeth S. Burnside, MD •
Margarita L. Zuley, MD • Robert D. Rosenberg, MD • Cindy S. Lee, MD
From the Departments of Radiology (W.A.B., M.L.Z.) and Computational and Systems Biology (J.M.B.), University of Pittsburgh School of Medicine, Pittsburgh, Pa;
Magee-Women’s Hospital of University of Pittsburgh Medical Center, 300 Halket St, Pittsburgh, PA 15213 (W.A.B., M.L.Z.); Department of Radiology and Biomedi-
cal Imaging, University of California San Francisco, San Francisco, Calif (E.A.S.); Department of Radiology, University of Wisconsin-Madison School of Medicine and
Public Health, Madison, Wis (E.S.B.); Radiology Associates of Albuquerque, Albuquerque, NM (R.D.R.); and Department of Radiology, New York University Langone
Medical Center, New York, NY (C.S.L.). Received November 30, 2019; revision requested January 2, 2020; revision received March 12; accepted March 25. Address
correspondence to W.A.B. (e-mail: [email protected]).
The views expressed in this article represent those of the author(s), and do not necessarily represent the official views of the American College of Radiology's National
Radiology Data Registry or the American College of Radiology.

Conflicts of interest are listed at the end of this article.

See also the editorial by Moy in this issue.

Radiology 2020; 296:32–41 • https://doi.org/10.1148/radiol.2020192641 • Content codes:

Background: The literature supports the use of short-interval follow-up as an alternative to biopsy for lesions assessed as probably be-
nign, Breast Imaging Reporting and Data System (BI-RADS) category 3, with an expected malignancy rate of less than 2%.

Purpose: To assess outcomes from 6-, 12-, and 24-month follow-up of probably benign findings first identified at recall from screen-
ing mammography in the National Mammography Database (NMD).

Materials and Methods: This retrospective study included women recalled from screening mammography with BI-RADS category 3 as-
sessment at additional evaluation from January 2009 through March 2018 from 471 NMD facilities. Only the first BI-RADS cat-
egory 3 occurrence for women aged 25 years or older with no personal history of breast cancer was analyzed, with biopsy or 2-year
imaging follow-up. Cancer yield and positive predictive value of biopsies performed (PPV3) were determined at each follow-up.

Results: Among 45 202 women (median age, 55 years; range, 25–90 years) with a BI-RADS category 3 lesion, 1574 (3.5%)
underwent biopsy at the time of lesion detection, yielding 72 cancers (cancer yield, 4.6%; 72 of 1574 women). For the remain-
ing 43 628 women who accepted surveillance, 922 were seen within 90 days (with 78 lesions biopsied and 12 [15%] classified
as malignant). The women still in surveillance (31 465 of 43 381 women [72.5%]) underwent follow-up mammography at 6
months. Of 3001 (9.5%) lesions biopsied, 456 (15.2%) were malignant (cancer yield, 1.5%; 456 of 31 465 women; 95% con-
fidence interval [CI]: 1.3%, 1.6%). Among 18 748 of 25 997 women (72.1%) in surveillance who underwent follow-up at 12
months, 1219 (6.5%) underwent biopsy with 230 (18.9%) malignant lesions found (cancer yield, 1.2%; 230 of 18 748 women;
95% CI: 1.1%, 1.4%). Through 2-year follow-up, the biopsy rate was 11.2% (4894 of 43 628 women) with a cancer yield of
1.86% (810 malignancies found among 43 628 women; 95% CI: 1.73%, 1.98%) and a PPV3 of 16.6% (810 malignancies
found among 4894 women).

Conclusion: In the National Mammography Database, Breast Imaging Reporting and Data System (BI-RADS) category 3 use is ap-
propriate, with 1.86% cumulative cancer yield through 2-year follow-up. Of 810 malignancies, 468 (57.8%) were diagnosed at or
before 6 months, validating necessity of short-interval follow-up of mammographic BI-RADS category 3 findings.
© RSNA, 2020

Online supplemental material is available for this article.

M ammographic screening helps reduce breast cancer

mortality because it can depict breast cancer before
it is clinically evident and usually before it has spread to
Reporting and Data System [BI-RADS] category 3, prob-
ably benign), with a less than 2% rate of malignancy in
multiple series (5–9). Avoiding initial biopsy of such le-
lymph nodes (1). In the United States, cancer detection sions reduces the harms and costs of screening.
rates average four to five per 1000 women per year (2–4) Mammographic lesions appropriately validated as
after recalling an average of 9.8%–11.6% of women for probably benign (BI-RADS category 3) include a solitary
additional testing. Most mammographic findings recalled circumscribed mass that appears solid at US (Fig 1), a focal
for additional testing prove to be normal tissue or benign asymmetry with no sonographic correlate, and a group of
findings; fewer than 2% prove suspicious and require bi- calcifications appearing punctate on magnification views
opsy, with 25%–30% of biopsies showing cancer (2–4). (5–9) (Fig 2). These findings were originally described on
For some specific findings, after initial diagnostic imag- baseline mammograms but also apply when prior com-
ing, short-term surveillance (usually in 6 months) can be parison images are unavailable. Importantly, a BI-RADS
safely recommended in lieu of biopsy (ie, Breast Imaging category 3 assessment should be made only after additional

This copy is for personal use only. To order printed copies, contact [email protected]
 Berg et al

With screening US, increasing data show that the rate of

Abbreviations malignancy for BI-RADS category 3 lesions is very low (0.4%–
BI-RADS = Breast Imaging Reporting and Data System, CI = confi- 0.8%); most cancers are not diagnosed until at least 12 months
dence interval, DCIS = ductal carcinoma in situ, NMD = National
Mammography Database, PPV3 = positive predictive value of biopsies of follow-up (15–17). The recommendation made in these pub-
performed lications is that BI-RADS category 3 lesions seen at screening
US can be followed up at the time of annual screening (ie, 12
Summary months rather than 6 months). Malignancy among sonographic
Of 43 628 women, 810 (1.86%) given Breast Imaging Reporting and BI-RADS category 3 masses is particularly unlikely in the ab-
Data System category 3 assessment after screening mammography
recall were diagnosed with malignancy, with 468 of 810 malignan- sence of a mammographic correlate (18).
cies (57.8%) diagnosed at or before 6 months, validating necessity of Prompted by these results from screening US, and by questions
short-interval follow-up. first raised in 1999 by Rubin (19) about the necessity of 6-month
Key Results follow-up, we sought to reevaluate the rate of malignancy at the
n Among 43 628 women undergoing surveillance after Breast
6-month and subsequent short-interval follow-up examinations
Imaging Reporting and Data System category 3 assessment in for BI-RADS category 3 after recall from screening mammog-
the National Mammography Database, cumulative cancer yields raphy. Our purpose was to assess outcomes from 6-, 12-, and
at 6, 12, and 24 months were 1.07%, 1.60%, and 1.86%, 24-month follow-up of probably benign findings first identified
respectively.
after recall from screening mammography in the NMD.
n For Breast Imaging Reporting and Data System category 3 find-
ings, 468 of 810 malignancies (57.8%) were diagnosed at or be-
fore 6-month follow-up. Materials and Methods
n When staging was available, 391 of 428 cancers (91.4%) were stage
0 or I with ductal carcinoma in situ overrepresented, without evi- This retrospective study was approved by the institutional review
dence of advanced stage due to the 6–24-month delay in diagnosis. board and compliant with the Health Insurance Portability and
Accountability Act. It included women recalled from screening
mammography with BI-RADS category 3 assessment at addi-
diagnostic work-up and not at screening mammography. Typi- tional evaluation from January 2009 through March 2018, from
cally, such lesions are recommended for 6-, 12-, and 24-month 471 NMD facilities in 31 states. We excluded duplicate records
follow-up and should receive a BI-RADS category 3 assessment (Fig 3). We limited our analysis to women with an indication for
at 6 and 12 months if stable (10,11). In addition to the requisite examination of screening who were given a BI-RADS category
low rate of malignancy for BI-RADS category 3 lesions, the brief 0 assessment (incomplete, needs additional imaging), followed
delay in diagnosis of the few cancers that are followed should not by a second visit (with indication for examination of recall from
harm the patient (12). The vast majority of such cancers should screening or diagnostic evaluation of breast problem) with a BI-
not have spread to lymph nodes or otherwise have adverse prog- RADS category 3 assessment within 90 days; then we followed
nosis at the time of diagnosis (13). up this population longitudinally. We analyzed only the first BI-
The National Mammography Database (NMD) was estab- RADS category 3 occurrences for women aged 25 years or older
lished in 2008 as part of the American College of Radiology with no reported personal history of breast cancer. The women
National Radiology Data
Registry. The primary
purpose of the NMD is
quality improvement: It
allows comparison of per-
formance of individual
radiologists and facilities
with their peers. Facilities
that decide to participate
can automatically up-
load data to the NMD
from the same software
used to meet Mammog-
raphy Quality Standards
Act requirements. Met-
rics parallel those of the
Breast Cancer Surveil-
lance Consortium (4),
although most NMD Figure 1: Circumscribed mass at baseline screening mammography in a 40-year-old woman due to a fibroadenoma. (a) Cranio-
metrics are lower because caudal tomosynthesis image shows a circumscribed mass (arrow), which was also evident on mediolateral oblique image (not shown).
of less complete cancer (b) Targeted transverse US image shows a circumscribed hypoechoic mass (arrow), a probably benign finding (Breast Imaging Re-
ascertainment (14). porting and Data System category 3). The patient preferred biopsy to surveillance, with US-guided core biopsy showing fibroadenoma.

Radiology: Volume 296: Number 1—July 2020 n radiology.rsna.org 33

Cancer Yield and Patterns of Follow-up for BI-RADS Category 3

Figure 2: Images in a 59-year-old woman with increasing calcifications at 6-month follow-up due to invasive and intraductal carcinoma. (a) Close-ups of mediolateral
oblique screening mammogram (left) and mediolateral oblique spot compression mammogram (right) show a few punctate and possibly layering calcifications (arrows).
Calcifications were not seen on magnification views, the standard craniocaudal projection, or prior mammograms (it is possible they were not included due to their posterior
location). Calcifications were assessed as Breast Imaging Reporting and Data System (BI-RADS) category 3, probably benign, with recommendation for 6-month follow-
up. (b) Spot magnification mediolateral (left) and craniocaudal (right) views obtained 7 months later show increase in the calcifications (arrows). Because of their increase,
stereotactic 9-gauge vacuum-assisted biopsy was performed, showing multiple 1–2-mm foci of nuclear grade 3 invasive ductal carcinoma with ductal carcinoma in situ
that was estrogen- and progesterone-receptor positive and human epidermal growth factor-2 receptor negative. Ki-67 proliferation index was 35%. There was no residual
invasive tumor at excision, and four sentinel nodes were negative. The short delay in diagnosis of the few cancers classified as BI-RADS category 3 does not increase likeli-
hood of regional spread to lymph nodes.

in the analysis set had visits at 369 of the 471 NMD facilities (c) breast density; (d) indication for examination (screening;
(78.3%). additional evaluation of recent screening mammogram; diag-
We limited our analysis set to women with proven benign nostic short-interval follow-up; diagnostic evaluation of breast
or malignant outcomes. To establish truth, women underwent problem; diagnostic, previous history of breast cancer; or un-
a percutaneous biopsy or imaging follow-up for at least 640 known); (e) BI-RADS assessment category for each visit; (f )
days; no results after 820 days were considered. A 6-month biopsy results (benign or malignant); (g) timing for most biop-
follow-up was defined as 91–270 days after BI-RADS cat- sies; and (h) availability of prior mammograms for comparison.
egory 3 assessment; 12-month as 271–460 days; 18-month as Tomosynthesis was performed at some facilities, but re-
461–640 days; and 24-month as 641–820 days. Biopsies were sults were not recorded separately from digital or even screen-
attributed to a given follow-up if they occurred within 90 days film mammography in the NMD in the time period of this
with no additional intervening imaging; this included 9187 of analysis; all are considered “mammograms” for this analysis.
9483 (96.9%) of all biopsies. Cancer yield was calculated as The initial data set included 19 675 062 mammographic ex-
number of breast cancers (invasive carcinoma or ductal car- aminations performed in 8 296 728 patients. The data set was
cinoma in situ [DCIS]) per number of women at each time restricted to 1 380 652 of 7 601 537 women (18.2%) recalled
point, and cumulatively at up to 2 years (820 days). Positive from screening, of whom 157 130 of 1 380 652 (11.4%) were
predictive value of biopsies performed (PPV3) was defined as given a BI-RADS category 3 assessment within 90 days (Fig
the number of women with cancer diagnosed per number of 3). Prior to analysis, we removed 56 589 duplicate visit re-
women with biopsy performed. We considered women to have cords. Of 108 811 women excluded due to inadequate fol-
previously undergone mammography if a record of such ex- low-up, 64 717 (59.5%) underwent imaging from 2016 to
isted within the NMD and assumed the examination was her 2018; thus, there was inadequate time to have had follow-up.
baseline if no prior mammograms were either recorded in the To avoid confounding by BI-RADS category 3 assessments at
NMD or listed as existing. follow-up, only the first occurrence of a BI-RADS category
Data obtained from NMD mammographic examinations 3 assessment was used to identify a woman for analysis. The
are recorded at the examination level (not lesion or breast “most significant mammographic finding” (mass, calcifica-
level) and include (a) masked patient identification, race, age, tion, architectural distortion, and asymmetry) was only re-
and year of examination; (b) masked facility identification; corded for 6124 of 43 628 women (14.0%) (2387 masses,

34 radiology.rsna.org n Radiology: Volume 296: Number 1—July 2020

 Berg et al

Results
There were 45 202 women (median age, 55 years; range, 25–90
years) recalled from screening mammography who underwent
additional imaging and were given their first BI-RADS category
3 assessment within 90 days (Table 1). For 1574 of these 45 202
women (3.5%), a biopsy was performed with their initial BI-
RADS category 3 assessment with 72 women found to have
malignancy (4.6%). The median age of these 1574 women, 52
years (range, 29–88 years), was lower than that of the remain-
ing 43 628 women or for the control group, and prior mammo-
grams were less likely to be available (P , .001 for each) (Table
1). These women likely differ from the usual BI-RADS category
3 population (women who accept surveillance) in other ways
and therefore were not considered further, leading to a study
population of 43 628 women and 168 798 mammograms (in-
cluding those obtained at initial additional imaging) for analysis.
Nine hundred twenty-two of 43 628 women (2.1%) were
seen for very short-interval follow-up within the first 90
days after a BI-RADS category 3 assessment. Of these, 78
(8.4%) underwent biopsy, and 12 of those 78 (15%) biop-
sies proved malignant (cancer yield, 12 of 922 [1.3%]). Im-
aging in only 169 of 922 women (18.3%) was downgraded
to BI-RADS category 1 or 2 at that visit (Table 2).

6-month Follow-up
Among 43 381 women remaining in surveillance, 31 465
Figure 3: Flowchart of analysis set. Among the 108 811 women excluded (72.5%) underwent 6-month follow-up imaging, including
because of insufficient follow-up were 64 417 (59.2%) who were examined 476 women who also underwent additional follow-up within
between 2016 and 2018 with insufficient time for follow-up. BI-RADS = Breast
the first 90 days but had neither undergone biopsy nor down-
Imaging Reporting and Data System.
grading to BI-RADS category 1 or 2. At the 6-month follow-
up, 14 383 of 31 465 women (45.7%) were downgraded to
2666 calcifications, 76 distortions, and 995 asymmetries) and BI-RADS category 1 or 2. Another 3001 of 31 465 women
was not further considered. (9.5%) underwent biopsy after 6-month follow-up, yielding
456 malignancies (PPV3, 15.2% [456 of 3001 women]; can-
Statistical Analysis cer yield, 1.5% [456 of 31 465 women]).
We compared cancer yields for women with BI-RADS cat- Completion of initial 6-month surveillance varied with
egory 3 assessments with those from a control group of patient age (Table 3). The percentage of women who had a
women with BI-RADS category 1 or 2 assessments at screen- follow-up visit at 6 months increased from 68.4% (798 of
ing in the NMD who had at least 640 days of follow-up or 1167) for women aged 30–39 years at initial BI-RADS cat-
biopsy within 820 days (n = 803 212). We also evaluated egory 3 assessment to 79.0% (3775 of 4780) for women aged
cancer yields within 820 days of initial BI-RADS category 70–79 years (test for trend P = .017).
3 assessment for women who were later downgraded to BI-
RADS category 1 or 2 and compared those to cancer yields 12-month Follow-up
for women still in surveillance for a BI-RADS category 3 Of 25 997 women remaining in surveillance (not yet having un-
assessment using the Fisher exact test. We also examined dergone downgrading or biopsy), 18 748 (72.1%) were seen for
the effect of patient age on completing 6- and 12-month follow-up at 12 months; of these, this was the first follow-up
follow-up. DCIS versus invasive tumor status was avail- for 7735 women (representing 17.7% of the 43 628 initial BI-
able for all malignancies; minimal cancer status (defined as RADS category 3 population). Of the 18 748 women, 11 897
invasive cancer 1 cm or smaller or DCIS) and nodal and (63.5%) were downgraded to BI-RADS category 1 or 2. An-
anatomic stage were assessed where available. A threshold other 1219 of the 18 748 women (6.5%) underwent biopsy after
of P = .05 was used for significance without adjustment for their 12-month follow-up, yielding 230 malignancies (PPV3,
multiple comparisons. We determined 95% confidence in- 18.9% [230 of 1219 women]; cancer yield, 1.2% [230 of 18 748
tervals (CIs) using 1.96 times the standard deviations cal- women]). After the 12-month time point, only 12 881 of 43 628
culated from Poisson statistics. All statistical analyses were women (29.5%) remained in surveillance (Fig 4). Excluding
performed using R software (version 1.1.383; RStudio, Bos- women aged 25–29 years, completion of 12-month follow-up
ton, Mass) (20,21). also increased with increasing patient age (Table 3).

Radiology: Volume 296: Number 1—July 2020 n radiology.rsna.org 35

Cancer Yield and Patterns of Follow-up for BI-RADS Category 3

Table 1: Demographics of Women with BI-RADS Category 3 Assessments versus Control Group in the NMD

P Value for P Value for

Excluded Owing Comparison of Comparison of
Study Population to Immediate Study Population vs Control Group Study Population
Demographic (n = 43 628)* Biopsy (n = 1574)† Excluded Women (n = 803 212)‡ vs Control Group
Age (y)
Median§ 55 (25–90) 52 (29–88) ,.001 59 (25–90) ,.001
,40 1176 (2.7) 72 (4.6) 3951 (0.5)
40–79 41 567 (95.3) 1467 (93.2) 772 279 (96.1)
80 885 (2.0) 35 (2.2) 26 982 (3.4)
Race and/or ethnicity|| .34 ,.001
White 16 006 (36.7) 591 (37.5) 325 759 (40.6)
Black 2604 (6.0) 71 (4.5) 42 737 (5.3)
Asian 509 (1.2) 17 (1.1) 8796 (1.1)
American Indian or Alaska Native 146 (0.3) 9 (0.6) 1630 (0.2)
Native Hawaiian or Pacific Islander 114 (0.3) 0 (0.0) 769 (0.01)
Other or missing 24 249 (55.6) 886 (56.3) 423 521 (52.7)
Breast density|| .086 ,.001
Fatty 1418 (5.3) 29 (4.9) 65 290 (9.6)
Scattered 12 639 (46.8) 256 (43.0) 326 731 (48.1)
Heterogeneously dense 11 682 (43.3) 265 (44.5) 251 069 (36.9)
Extremely dense 1266 (4.7) 46 (7.7) 36 546 (5.4)
Missing 16 623 (38.1) 978 (62.1) 123 576 (15.4)
Prior mammograms available|| ,.001 ,.001
Yes 25 160 (57.7) 643 (40.9) 487 406 (60.7)
No 18 468 (42.3) 931 (59.1) 315 806 (39.3)
Note.—Except where indicated, data are numbers of women, with percentages in parentheses. BI-RADS = Breast Imaging Reporting and
Data System, NMD = National Mammography Database.
* A total of 1574 women who underwent immediate biopsy at the time of their BI-RADS 3 assessment were excluded from the study population.
†
Six hundred forty-one of the 1574 women (40.7%) who underwent immediate biopsy were seen at one of three facilities, suggesting dif-
ferences in practice.
‡
The control group consisted of women with BI-RADS category 1 or 2 assessments at screening and at least 640 days of follow-up or
biopsy within 820 days.
§
Numbers in parentheses are the range. The same median age of 55 years and range of 25–90 years also applies to the entire group of
45 202 (43 628 + 1574) women.
||
Percentages given use as denominator those where results were available.

18-month Follow-up this time point (representing 2317 of 43 628 [5.3%] of the initial
Only 4742 of 12 881 women (36.8%) remaining in surveillance, BI-RADS category 3 population). Of the 7352 women seen at
of whom 4401 were still classified as having BI-RADS category 24-month follow-up, 256 (3.5%) underwent biopsy, yielding 42
3 lesions, were seen for follow-up at 18 months; 70 cancers were malignancies (PPV3, 16.4%; cancer yield, 0.6%); 6115 women
diagnosed (cancer yield, 1.5%; 95% CI: 1.1%, 1.8%). For 1665 (83.2%) were downgraded to BI-RADS category 1 or 2. Only
women, this was their first follow-up (representing 3.8% of the 3409 of 43 628 (7.8%) of the original population remained in
43 628 initial BI-RADS category 3 population). Among those surveillance after 2 years. Figure 5 shows cumulative malignancy
seen for 18-month follow-up were 1504 women who had also rates at follow-up over time, both before and after downgrade to
been seen at 12 months (lesions remained BI-RADS category 3), BI-RADS category 1 or 2. Overall biopsy rate after 2 years for
with 18 malignancies among 110 biopsies in this subset (cancer women with BI-RADS category 3 assessments was 11.2% (4894
yield, 1.2%; 95% CI: 0.6%, 1.8%). A subset of 1205 of the of 43 628 women), with a cancer yield of 1.86% (810 of 43 628
women seen at 18 months had been seen for initial follow-up at women; 95% CI: 1.73%, 1.98%) and a PPV3 of 16.6% (810
6 months, and there were 17 malignancies among 86 biopsies in of 4894 women). Interestingly, only 2648 of 43 628 women
this group (cancer yield, 1.4%; 95% CI: 0.7%, 2.1%). (6.07%) underwent the complete set of 6-, 12-, and 24-month
follow-up examinations.
24-month Follow-up
Of 9780 women remaining in surveillance through 18 months Staging
of follow-up, 7352 (75.2%) were seen for 24-month follow-up, Of 810 malignancies, 265 (32.7%) were DCIS, and, where
and 2317 of these women were seen for their first follow-up at nodal status was available, only 27 of 276 (13.4%) invasive

36 radiology.rsna.org n Radiology: Volume 296: Number 1—July 2020

 Berg et al

Table 2: Summary of Follow-up of Women with Lesions Remaining as BI-RADS Category 3

No. of Women
with Imaging
Interval Since Initial No. of Women Cancer Yield Downgraded to
BI-RADS 3 Assessment No. of Women with Visit No. of Women No. of per Time BI-RADS
after Screening Recall in Surveillance Recorded with Biopsy Cancers PPV3 (%) Period (%)* Category 1 or 2
0–90 days† 43 628 922 78 (8.5) 12 15 1.3 (0.6, 2.0) 169
6 months (91–270 d) 43 381 31 465 (72.5)‡ 3001 (9.5) 456 15.2 1.5 (1.3, 1.6) 14 383
12 months (271–460 d) 25 997 18 748 (72.1)† 1219 (6.5) 230 18.9 1.2 (1.1, 1.4) 11 897
18 months (461–640 d) 12 881 4742 (36.8)† 340 (7.2) 70 20.6 1.5 (1.1, 1.8) 2761
24 months (641–820 d) 9780 7352 (75.2)† 256 (3.5) 42 16.4 0.6 (0.4, 0.7) 6115
Overall 43 628 4894 (11.2) 810 16.6 1.86 (1.73, 1.98) 35 325
Note.—Except where indicated, data are numbers of women, with percentages in parentheses. BI-RADS = Breast Imaging Reporting and
Data System, PPV3 = positive predictive value of biopsies performed.
* Cancer yield is the number of cancers diagnosed in that interval divided by number of women seen for follow-up in that interval. Num-
bers in parentheses are 95% confidence intervals.
†
For another 1574 women, a biopsy was associated with the initial BI-RADS category 3 assessment at day 0, with 72 (4.6%) malignant.
These women were excluded from analysis of surveillance.
‡
This includes a small number of women with imaging reclassified as BI-RADS category 0, 4, or 5 after the prior follow-up visit: 22
women at 6 months, 389 women at 12 months, 341 women at 18 months, and 653 women at 24 months.

cancers were node positive (Table 4). Complete staging infor- downgraded imaging (P , .001 for all comparisons). The per-
mation was available for 428 of 810 malignancies (52.8%) in centage of malignancies that were DCIS (32.7%, 265 of 810)
this data set (Table 4), including 265 DCIS (stage 0) and 126 was higher for BI-RADS category 3 cancers than for those di-
stage I invasive malignancies (ie, of the 428 with staging, 391 agnosed after downgrade (20.4%, 37 of 181) or those in the
[91.4%] were stage 0 or I). Of the 507 malignancies with status control group (23.7%, 821 of 3468) (P = .001 and P  .001,
recorded, 389 (76.7%) were minimal cancers. Among another respectively). The percentage that was minimal (76.7%, 389 of
181 cancers diagnosed during a similar 820-day interval in 507) was higher for BI-RADS category 3 cancers than for those
women with imaging downgraded to BI-RADS category 1 or diagnosed after downgrade (62.6%, 72 of 115) or those in the
2, staging was available for 90 (49.7%); 76 of the 90 cancers control group (66.1%, 1522 of 2304) (P = .003 and P , .001,
(84.4%) were stage 0 or I (37 stage 0 and 39 stage I). respectively), with no difference between downgraded and con-
trol groups (P = .48). The rate of node-negative invasive cancers
Outcomes after Downgrade versus Control Group with BI- with node staging (13.4%, 37 of 276) was also lower than that
RADS Category 1 or 2 Assessment at Imaging after downgrade (17.2%, 15 of 87) or that in the control group
Imaging in 14 552 women was downgraded to BI-RADS cat- (15.5%, 211 of 1361), but differences were not significant (P =
egory 1 or 2 through 6 months of follow-up; 288 of those .38 and P = .41, respectively).
14 552 women (1.98%) underwent biopsy within the ensu-
ing 12 months, yielding 75 malignancies (cancer yield, 0.5%; Discussion
95% CI: 0.4%, 0.6%) (Table E1 [online], Fig 5). Imaging in In the National Mammography Database, we found use of
another 11 897 women was downgraded to BI-RADS category Breast Imaging Reporting and Data System (BI-RADS) cat-
1 or 2 at the 12-month follow-up examination; 177 of those egory 3 assessments after recall from screening mammography
11 897 women (1.5%) underwent biopsy within 12 months is appropriate, with 1.86% cumulative cancer yield after 820
of downgrade, yielding 65 malignancies (cancer yield, 0.6%; days. Rate of 6-month follow-up was relatively high at 72.5%
95% CI: 0.4%, 0.7%), a cumulative biopsy rate of 1.8% (465 (31 465 of 43 381 women), although 17.7% (7735 of 43 628
of 26 449), 140 cancers, and cancer yield of 0.5% (95% CI: women) did not undergo their first follow-up until 12 months.
0.4%, 0.6%). By comparison, among 803 212 women in the Of 810 cancers, 468 (57.8%) were diagnosed at or before the
control group with BI-RADS category 1 or 2 assessments, 6-month visit, validating the need for short-interval (usually
5458 (0.7%) underwent biopsy within the ensuing 12 months, 6-month) follow-up. Short-interval follow-up imaging surveil-
yielding 1326 malignancies (cancer yield, 0.17%; 95% CI: lance is further supported by malignancy rates substantially
0.16%, 0.17%; P , .001). exceeding those of women with imaging downgraded to BI-
Women with lesions remaining BI-RADS category 3 had RADS category 1 or 2 at each follow-up and exceeding those
468, 698, and 810 cancers diagnosed by 6 months, 12 months, of a control group of women given BI-RADS category 1 or 2
and 24 months, corresponding to cumulative cancer yields of assessments.
1.1%, 1.6%, and 1.86%, respectively. All of these cancer yields The more than 72% completion rate for 6-month follow-up
are significantly higher than the 0.6% rate for women with we observed is similar to that in prior reports. Bowles et al (22)

Radiology: Volume 296: Number 1—July 2020 n radiology.rsna.org 37

Cancer Yield and Patterns of Follow-up for BI-RADS Category 3

Table 3: Rates of Completion of 6- and 12-month Follow-up as a Function of Patient Age

Percentage of
No. of Women No. of Women Rate of No. of Original Popu- No. of Women Rate of
with BI-RADS with Follow-up Completion Women Still in lation Still in with Follow-up Completion
Category 3 at 6 Months of 6-month Surveillance at Surveillance at at 12 Months at 12-month
Age Group Assessment (0–270 d) Follow-up (%) 12 Months 12 Months (271–460 d) Follow-up (%)
25–29 y 9 5 55.6 6 66.7 6 100
30–39 y 1167 798 68.4 766 65.6 476 62.1
40–49 y 13 578 9313 68.6 8608 63.4 5965 69.3
50–59 y 12 687 8953 70.6 7670 60.5 5373 70.1
60–69 y 10 522 7929 75.4 5939 56.4 4524 76.2
70–79 y 4780 3775 79.0 2510 52.5 2006 79.9
80–90 y 885 692 78.2* 498 56.3* 398 79.9*
All ages 43 628 31 465 72.1 25 997 59.6 18 748 72.1
Note.—BI-RADS = Breast Imaging Reporting and Data System.
* Test for trend, P = .017 (not including point for women aged 25–29 years owing to the small number of women).

we only analyzed women who completed follow-up

within at least 2 years; thus, our patient population is
more compliant than average. Indeed, survey of the
same NMD data set for all women with a first occur-
rence of BI-RADS category 3 prior to 2016 and not
constraining for further follow-up revealed that only
61 535 of 113 070 of women (54.4%) underwent
6-month follow-up.
We limited our analysis to women recalled
from screening mammography who under-
went additional imaging. Although a BI-RADS
category 3 assessment can be given directly at
screening US (assuming orthogonal images have
been obtained) or at screening MRI, it should
only be used for mammographic findings after
Figure 4: Graph shows timing of resolution of a Breast Imaging Reporting and Data System (BI-
full diagnostic work-up (12). More advanced-
RADS) category 3 assessment (number of the initial 43 628 women who remained BI-RADS cat- stage distribution has been shown for cancers
egory 3 as a function of time, solid line). After 12-month follow-up, only 12 881 women (29.5%) after BI-RADS category 3 assessment directly
remained in surveillance. Overall, 4894 women (11.2%) underwent biopsy (dashed line), and at screening mammography (24), and women
imaging for 35 325 women (81.0%) was downgraded to BI-RADS category 1 or 2 (dotted line) may experience unnecessary stress from delay in
after 2 years of follow-up.
diagnostic work-up if recommended for surveil-
lance for what could have been shown to be a
showed that 90.9% of 15 515 diagnostic mammograms assigned simple cyst at immediate additional imaging. It is a pay-
a BI-RADS category 3 assessment were given a recommendation for-performance metric from the U.S. Centers for Medicare
for short-interval follow-up, 4.3% for normal follow-up, 3.0% for and Medicaid Services not to use BI-RADS category 3 at
additional imaging, and 1.8% for surgical consultation or biopsy. screening mammography.
Of 398 women recommended for 6-month follow-up after mam- We do not know what prompted biopsy at initial diagnos-
mography, US, or both, 233 (58.5%) were followed up within tic work-up, and the cancer detection rate of 4.6% (72 of 1574
5 months of the recommended interval in a study by Borders et women) in that cohort is considerably higher than the 1.86%
al (23). In addition, the women who complied with follow-up malignancy rate observed otherwise from a BI-RADS category 3
recommendations reported that mailed reminder cards were more assessment (P , .0001). This population may be distinct from
helpful than telephone reminders. In that series, women who did other BI-RADS category 3 patients with some lesions misclassi-
not comply with recommendations for follow-up were younger fied, knowing that biopsy would be performed. These patients
than those who complied with follow-up (mean age, 49 years vs 54 were younger and may also have been at higher risk (with higher
years, respectively). We similarly found reduced 6- and 12-month pretest probability) for cancer and therefore less inclined to ac-
follow-up rates among younger women in this series. Follow-up cept surveillance in lieu of biopsy.
limited to short-interval US alone would not be captured by our We also observed that some women underwent initial
analysis; thus, compliance may be underestimated. Conversely, follow-up or even biopsy less than 90 days after initial

38 radiology.rsna.org n Radiology: Volume 296: Number 1—July 2020

 Berg et al

Table 4: Staging of Cancers Diagnosed at Surveillance of BI-RADS Category 3 Lesions in the NMD

No. of
No. of Minimal No. of Cancers
No. of Cancers/No. of No. of No. with Node- No. of No. of No. of No. of with
No. of DCIS Reported Invasive Node Positive Stage I Stage II Stage III Stage IV Stage
Parameter Cancers Lesions Cancers* Cancers Staging† Cancers‡ Cancers§ Cancers Cancers Cancers Missing
BI-RADS Category 3
0–90 days 12 2 (16.7) 2/7 (29) 10 6 (60) 1 (16.7) 2 1 NA NA 7
6 months 456 165 (36.2) 242/296 (82.0) 291 147 (50.5) 18 (12.2) 64 12 2 1 212
  (91–270 d)
12 months 230 67 (29.2) 101/141 (71.6) 163 82 (50) 10 (12.2) 41 10 3 NA 109
  (271–460 d)
18 months 70 21 (30.0) 29/41 (70.7) 49 26 (53) 7 (26.9) 9 3 1 NA 36
  (461–640 d)
24 months 42 10 (23.8) 15/23 (65) 32 15 (47) 1 (6.7) 10 4 NA NA 18
  (641–820 d)
Cumulative 810 265 (32.7) 389/507 (76.7) 545 276 (50.6) 37 (13.4) 126|| 30 6 1 382
   BI-RADS
category 3
Downgrade to BI-RADS Category 1 or 2#
0–90 days 1 0 (0.0) NA 1 0 (0.0) 0 2 NA NA NA 1
6 months 113 22 (19.5) 40/70 (57) 91 54 (59.3) 8 (14.8) 24 7 NA NA 60
  (91–270 d)
12 months 65 14 (21.5) 30/43 (70) 51 32 (62.7) 6 (18.8) 12 5 1 1 32
  (271–460 d)
18 months 2 1 (50.0) 2/2 (100) 1 1 (100.0) 0 (0.0) 1 NA NA NA 1
  (461–640 d)
Cumulative 181 37 (20.4) 72/115 (62.6) 144 87 (60.4) 15 (17.2) 39** 12 1 1 94
   BI-RADS
category 1
or 2
Control Group BI-RADS Category 1 or 2
0–460 days 1369 318 (23.2) 606/936 (64.7) 1051 559 (53.2) 88 (15.7) 206 86 10 1 748
461–820 days 2099 503 (24.0) 916/1368 (67.0) 1596 802 (50.3) 123 (15.3) 320 87 18 2 1169
Cumulative 3468 821 (23.7) 1522/2304 (66.1) 2647 1361 (51.4) 211 (15.5) 526†† 173 28 3 1917
   BI-RADS
category 1
or 2
Note.—Data are numbers of cancers, with percentages in parentheses. BI-RADS = Breast Imaging Reporting and Data System, DCIS =
ductal carcinoma in situ, NA = not applicable, no entries, NMD = National Mammography Database.
* Minimal cancer is defined as invasive breast cancer 1 cm or smaller or DCIS; denominator is number of cancers where minimal (yes/no)
is reported.
†
Data are the numbers of invasive cancers with node staging.
‡
Percentage refers to invasive cancers that are node positive as a fraction of those with node staging.
§
Anatomic stage.
||
Of those with staging (ie, 265 + 126 + 30 + 6 + 1 = 428), a total of 391 (265 + 126) were stage 0 or I (ie, 391 of 428 [91.4%]).
#
Staging on cancers diagnosed as a function of timing of the downgrade, considering overall follow-up of 820 days since the original BI-
RADS category 3 assessment. Timing of diagnoses with cancer is detailed in Table E1 (online).
** Of those with staging (ie, 37 + 39 + 12 + 1 + 1 = 90), a total of 76 (37 + 39) were stage 0 or I (ie, 76 of 90 [82%]).
††
Of those with staging (ie, 821 + 526 + 173 + 28 +3 = 1551), a total of 1347 (821 + 526) were stage 0 or I (ie, 1347 of 1551 [86.8%]).

BI-RADS category 3 assessment. Women with suspected The BI-RADS guidance chapter suggests 6-, 12-, and
breast abscesses or fat necrosis are sometimes given recom- 24-month follow-up for BI-RADS category 3 findings stable at
mendations for very short-interval follow-up in a matter of each of the 6- and 12-month examinations (12); only 2648 of
weeks, particularly when the clinical scenario is appropriate. 43 628 women (6.1%) given a BI-RADS category 3 assessment
We do not have details of findings in this patient subpopu- after screening recall actually had this full sequence of examina-
lation nor do we know what prompted biopsy in general. tions in this analysis. It is expected that lesions decreasing or

Radiology: Volume 296: Number 1—July 2020 n radiology.rsna.org 39

Cancer Yield and Patterns of Follow-up for BI-RADS Category 3

Consortium, where 76.9% of cancers were stage

0 or I (2). Linkage of NMD data to state tumor
registries should improve collection of staging and
also recognition of false-negatives, and this has been
shown to be feasible (28).
Our study has several limitations. We do not
have a description of most of the mammographic or
tomosynthesis findings (eg, mass, calcifications, dis-
tortion, or asymmetry) prompting recall or biopsy,
nor do we know what work-up was performed. We
do not know how many women underwent tomo-
synthesis, which has been shown to slightly decrease
use of BI-RADS category 3 assessments (29,30).
Because the NMD does not provide lesion- or
breast-level outcomes, it is possible that some can-
cers diagnosed do not correspond to the lesions
Figure 5: Graph shows cumulative malignancy rate over time among 43 628 women with
assessed as BI-RADS category 3. We assumed that
Breast Imaging Reporting and Data System (BI-RADS) category 3 lesions at recall from screening
mammography. Through 2 years of follow-up, cancer was diagnosed in 810 women (1.86%, solid the next biopsy after a screening mammogram was
line). This rate is less than the 2% malignancy rate expected with a BI-RADS category 3 assessment. related to the abnormality seen at screening, but it
Malignancy rate among women with lesions downgraded to BI-RADS category 1 or 2 was 0.54% could have been a new finding elsewhere in either
within 12 months of downgrade (dotted line). breast. Our results apply only to screening and not
to palpable masses given a BI-RADS category 3 as-
resolved at any time point would be downgraded to BI-RADS sessment or to women with a personal history of breast cancer.
category 2 or 1, respectively. Differences can also exist in radi- In conclusion, across 43 628 women at 369 facilities in
ologists’ interpretations, and this is particularly an issue for BI- the National Mammography Database, use of Breast Imaging
RADS category 3 assessments (25). It is interesting to note that Reporting and Data System category 3 for findings recalled
1504 of 12 881 women (11.7%) with lesions that remained BI- from screening mammography appears appropriate, with 810
RADS category 3 after 12-month follow-up underwent follow- women diagnosed with cancer: The overall malignancy rate
up again at 18 months. Of those 1504 women, 110 (7.3%) through 2 years of 1.86% is below the expected rate of 2%. In
then underwent biopsy, yielding 18 malignancies (cancer yield, our study, 72.5% of women assessed as having probably benign
1.2% [18 of 1504]), with biopsy and malignancy rates not findings underwent 6-month follow-up mammography, and
significantly different from those at the 12-month follow-up 468 of 810 malignancies (57.8%) were diagnosed at or before
(6.4% [1184 of 18 359, P = .19] and 1.2% [223 of 18 359, P 6 months, validating the necessity of short-interval follow-up.
. .99], respectively). The 18 malignancies detected due to the
“extra” 18-month follow-up in this analysis represent 2.2% (18 Acknowledgements: This research was supported by the American College of
Radiology’s National Radiology Data Registry (NRDR). The authors wish to thank
of 810) of all malignancies. Buch et al (26) reviewed records for ACR staff for assistance in preparation of the National Mammography Database
7632 women with BI-RADS category 3 assessments undergo- (NMD) data and acknowledge guidance and input by the NMD steering commit-
ing 6-, 12-, 18-, and 24-month follow-up or biopsy and deter- tee for this analysis.

mined that 179 women ultimately underwent biopsy, yielding

Author contributions: Guarantors of integrity of entire study, W.A.B., J.M.B.,
34 malignancies. However, only 11 of the 179 (6.1%) biopsies C.S.L.; study concepts/study design or data acquisition or data analysis/interpreta-
were prompted by change at 18 months, with diagnosis of only tion, all authors; manuscript drafting or manuscript revision for important intel-
three of 34 (8.8%) malignancies. They concluded that 6, 12, lectual content, all authors; approval of final version of submitted manuscript, all
authors; agrees to ensure any questions related to the work are appropriately re-
and 24-month follow-ups were sufficient. solved, all authors; literature research, W.A.B., E.A.S., E.S.B., C.S.L.; clinical stud-
We do not have subtype information on cancers. Sickles (6), ies, W.A.B., E.A.S., E.S.B., C.S.L.; experimental studies, E.A.S., E.S.B.; statistical
Varas et al (8,27), and Vizcaino et al (9) have previously found analysis, J.M.B.; and manuscript editing, all authors
that rates of node positivity are lower among cancers detected
Disclosures of Conflicts of Interest: W.A.B. disclosed no relevant relationships.
after initial short-interval follow-up compared with those seen J.M.B. disclosed no relevant relationships. E.A.S. disclosed no relevant relation-
at screening. We only have staging for 428 of the 810 cancers ships. E.S.B. disclosed no relevant relationships. M.L.Z. Activities related to the
(52.8%) in this series, and 391 of 428 (91.4%) were stage 0 or present article: disclosed no relevant relationships. Activities not related to the pres-
ent article: receives reimbursement for board travel expenses from the Society of
stage I. The overall high rates of DCIS (32.7%), minimal cancer Breast Imaging; institution has grants/grants pending with Hologic. Other relation-
(76.7%), and node-negative invasive cancers (86.6%) we ob- ships: disclosed no relevant relationships. R.D.R. disclosed no relevant relation-
served among BI-RADS category 3 cancers suggest more favor- ships. C.S.L. disclosed no relevant relationships.

able prognosis tumors; however, 19 of the 153 invasive cancers

(12%) diagnosed within 6 months with node staging were node References
positive, indicating the need for close surveillance rather than 1. Smith RA, Duffy SW, Gabe R, Tabar L, Yen AM, Chen TH. The randomized trials of
breast cancer screening: what have we learned? Radiol Clin North Am 2004;42(5):793–
follow-up at 1 year. Our results compare favorably to staging 806, v.
seen for all women screened in the Breast Cancer Surveillance

40 radiology.rsna.org n Radiology: Volume 296: Number 1—July 2020

 Berg et al

2. Lehman CD, Arao RF, Sprague BL, et al. National Performance Benchmarks for Mod- 18. Chae EY, Cha JH, Shin HJ, Choi WJ, Kim HH. Reassessment and Follow-Up Re-
ern Screening Digital Mammography: Update from the Breast Cancer Surveillance sults of BI-RADS Category 3 Lesions Detected on Screening Breast Ultrasound. AJR
Consortium. Radiology 2017;283(1):49–58. Am J Roentgenol 2016;206(3):666–672.
3. Rosenberg RD, Yankaskas BC, Abraham LA, et al. Performance benchmarks for 19. Rubin E. Six-month follow-up: an alternative view. Radiology 1999;213(1):15–18;
screening mammography. Radiology 2006;241(1):55–66. discussion 19–21.
4. Lee CS, Bhargavan-Chatfield M, Burnside ES, Nagy P, Sickles EA. The National Mam- 20. RStudio Team. RStudio: Integrated development for R. Boston, Mass: RStudio, 2015.
mography Database: Preliminary Data. AJR Am J Roentgenol 2016;206(4):883–890. 21. Wickham H, Francois R, Henry L, Müller K. dplyr: A Grammar of Data Manipula-
5. Sickles EA. Breast calcifications: mammographic evaluation. Radiology 1986;160(2): tion. 2018.
289–293. 22. Bowles EJ, Sickles EA, Miglioretti DL, Carney PA, Elmore JG. Recommendation for
6. Sickles EA. Periodic mammographic follow-up of probably benign lesions: results in short-interval follow-up examinations after a probably benign assessment: is clinical
3,184 consecutive cases. Radiology 1991;179(2):463–468. practice consistent with BI-RADS guidance? AJR Am J Roentgenol 2010;194(4):
7. Sickles EA. Management of probably benign breast lesions. Radiol Clin North Am 1152–1159.
1995;33(6):1123–1130. 23. Borders MH, Cheng L, Fitzpatrick KA, Krupinski EA. Patient Compliance in the
8. Varas X, Leborgne JH, Leborgne F, Mezzera J, Jaumandreu S, Leborgne F. Revisiting Setting of BI-RADS Category 3: What Factors Impact Compliance With Short-
the mammographic follow-up of BI-RADS category 3 lesions. AJR Am J Roentgenol Term Follow-Up Recommendations? Breast J 2017;23(1):77–82.
2002;179(3):691–695. 24. Kerlikowske K, Smith-Bindman R, Abraham LA, et al. Breast cancer yield for screen-
9. Vizcaíno I, Gadea L, Andreo L, et al. Short-term follow-up results in 795 non- ing mammographic examinations with recommendation for short-interval follow-
palpable probably benign lesions detected at screening mammography. Radiology up. Radiology 2005;234(3):684–692.
2001;219(2):475–483. 25. Berg WA, Campassi C, Langenberg P, Sexton MJ. Breast Imaging Reporting and
10. Sickles EA. Probably benign breast lesions: when should follow-up be recommended Data System: inter- and intraobserver variability in feature analysis and final assess-
and what is the optimal follow-up protocol? Radiology 1999;213(1):11–14. ment. AJR Am J Roentgenol 2000;174(6):1769–1777.
11. Sickles EA, D’Orsi CJ, Bassett LW, et al. ACR BI-RADS Mammography. ACR BI- 26. Buch KA, Qureshi MM, Carpentier B, et al. Surveillance of probably benign (BI-
RADS Atlas, Breast Imaging Reporting and Data System. Reston, Va: American Col- RADS 3) lesions in mammography: what is the right follow-up protocol? Breast J
lege of Radiology, 2013. 2015;21(2):168–174.
12. Sickles EA, D’Orsi CJ. Follow-up and outcome monitoring. ACR BI-RADS Atlas, 27. Varas X, Leborgne F, Leborgne JH. Nonpalpable, probably benign lesions: role of
Breast Imaging Reporting and Data System. Reston, Va: American College of Radi- follow-up mammography. Radiology 1992;184(2):409–414.
ology, 2013. 28. Zuley ML, Nishikawa RM, Lee CS, et al. Linkage of the ACR National Mammog-
13. Badve SS, Beitsch PD, Bose S, et al. Part XI Breast. AJCC Cancer Staging Manual. raphy Database to the Network of State Cancer Registries: Proof of Concept Evalu-
8th ed. Chicago, Ill: American College of Surgeons, 2018. ation by the ACR National Mammography Database Committee. J Am Coll Radiol
14. D’Orsi CJ, Sickles EA. 2017 Breast Cancer Surveillance Consortium Reports on In- 2019;16(1):8–14.
terpretive Performance at Screening and Diagnostic Mammography: Welcome New 29. McDonald ES, McCarthy AM, Weinstein SP, Schnall MD, Conant EF. BI-RADS Cate-
Data, But Not as Benchmarks for Practice. Radiology 2017;283(1):7–9. gory 3 Comparison: Probably Benign Category after Recall from Screening before and af-
15. Barr RG, Zhang Z, Cormack JB, Mendelson EB, Berg WA. Probably benign lesions ter Implementation of Digital Breast Tomosynthesis. Radiology 2017;285(3):778–787.
at screening breast US in a population with elevated risk: prevalence and rate of 30. Stepanek T, Constantinou N, Marshall H, et al. Changes in the Utilization of the BI-
malignancy in the ACRIN 6666 trial. Radiology 2013;269(3):701–712. RADS Category 3 Assessment in Recalled Patients Before and After the Implemen-
16. Nam SY, Ko EY, Han BK, Shin JH, Ko ES, Hahn SY. Breast Imaging Reporting and tation of Screening Digital Breast Tomosynthesis. Acad Radiol 2019;26(11):1515–
Data System Category 3 Lesions Detected on Whole-Breast Screening Ultrasound. J 1525.
Breast Cancer 2016;19(3):301–307.
17. Moon HJ, Kim MJ, Yoon JH, Kim EK. Follow-up interval for probably benign
breast lesions on screening ultrasound in women at average risk for breast cancer with
dense breasts. Acta Radiol 2018;59(9):1045–1050.

Radiology: Volume 296: Number 1—July 2020 n radiology.rsna.org 41