OBJECTIVE. The objective of this study was to prospectively compare dy
namic contrast-enhanced MR imaging with MR sequences previously descri
bed for assessing perianal fistulas in order to determine the best MR
protocol for their evaluation. SUBJECTS AND METHODS. MR examinations o
f 42 consecutive patients with clinically suspected perianal fistulas
were independently evaluated by two experienced observers blinded to t
he findings of digital rectal examination. The observers' evaluations
occurred before definitive surgical exploration. All patients had body
-coil MR imaging examinations, including the following sequences that
were ranked for anatomic and pathologic information: spin-echo T1-weig
hted, short inversion time inversion recovery, and dynamic contrast-en
hanced MR imaging in the coronal plane; and spin-echo T2-weighted imag
ing in the axial plane. Surgical findings were accepted as the gold st
andard and were recorded independently by the surgeon, who was unaware
of the findings of the MR assessment. MR findings were subsequently c
orrelated with digital rectal examination before surgery and with clin
ical follow-up. RESULTS. MR imaging correctly allowed our blinded obse
rvers to predict the surgical anatomy of perianal disease in 37 of the
42 patients (accuracy, 88%). For detection of the presence and site o
f an enteric fistulous entry, MR imaging had a sensitivity of 97%, a s
pecificity of 67%, a positive predictive value of 88%, and a negative
predictive value of 89%. On MR imaging examination, eight patients had
no fistula, 12 had simple intersphincteric fistulas, and 22 had compl
ex fistulas. MR imaging revealed all 14 perianal abscesses and fluid c
ollections found at surgery. Digital rectal examination before surgery
failed to reveal abscesses or important secondary tracks in eight of
the 22 complex fistulas. For anatomic and pathologic depiction of fist
ulas, dynamic contrast-enhanced MR imaging ranked as the best sequence
for 22 of 34 fistulas. The short inversion time inversion recovery se
quence, which was unable to distinguish small abscesses from perianal
inflammation and showed spurious high signal in old fibrotic tracks, l
ed our observers to misdiagnose five cases. In four patients for which
initial surgery did not confirm enteric entry sites that our observer
s had predicted by MR imaging, follow-up has confirmed the observers'
diagnoses. The observers' evaluations of the MR examinations agreed in
37 (88%) of the 42 cases. CONCLUSION. MR imaging is more accurate tha
n digital rectal examination before surgery in detecting complex featu
res of perianal fistulas. MR imaging is noninvasive, is highly accurat
e, and has low interobserver variability. With MR imaging, observers m
ay better predict outcome than with initial surgical exploration. MR a
ssessment that includes dynamic contrast-enhanced MR imaging and axial
T2-weighted sequences (examination time, 20 min) provides the anatomi
c and pathologic information required to guide surgical management.