European Journal of Cancer 38 (2002) 887–898

www.ejconline.com

Radiology of colorectal cancer

M.E.J. Pijla,*, A.S. Chaouib, R.L. Wahlc, J.A. van Oostayena
a
Department of Radiology, Leiden University Medical Centre, Albinusdreef 2, 2333 ZA Leiden, The Netherlands
b
Department of Radiology, Boston Medical Center, Boston University School of Medicine, Boston, MA, USA
c
Department of Nuclear Medicine, Johns Hopkins Medical Institute, Baltimore, MD, USA

Received 12 December 2001; accepted 6 February 2002

Abstract
In the past 20 years, the radiology of colorectal cancer has evolved from the barium enema to advanced imaging modalities like
phased array magnetic resonance imaging (MRI), virtual colonoscopy and positron emission tomography (PET). Nowadays, pri-
mary rectal cancers are preferably imaged with transrectal ultrasound or MRI, while barium enema is still the most often used
technique for imaging of colonic cancers. Virtual colonoscopy is rapidly evolving and might considerably change the imaging of
colorectal cancer in the near future. The use of virtual colonoscopy for screening purposes and imaging of the colon in occlusive
cancer or incomplete colonoscopies is currently under evaluation. The main role of PET is in detecting tumour recurrences, both
locally and distantly. Techniques to fuse cross-sectional anatomical (computer tomography (CT) and MRI) and functional (PET)
images are being developed. Apart from diagnostic imaging, the radiologists has added image-guided minimally invasive treatments
of colorectal liver metastases to their arsenal. The radio-frequency ablation technique is now widely available, and can be used
during laparotomy or percutaneously in selected cases. # 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Colorectal; Review; Radiology; Diagnostic imaging

1. Radiology of colorectal cancer multi-detector CT and PET can also be used to image
the direct extension, lymphatic spread and peritoneal
Radiology of colorectal cancer has dramatically seeding of colorectal cancer. Image-guided, minimally
evolved over the past two decades. In the 1980s, colo- invasive treatment of colorectal liver metastases is
rectal cancer could be diagnosed with a barium enema, another new component in the radiology of colorectal
and its liver metastases with ultrasound (US) and incre- cancer.
mental computer tomography (CT). Nowadays, there is The radiology of colorectal cancer is addressed in this
a definite difference in the imaging approach of either review in four separate sections. In the first section, the
colonic or rectal cancer. The method of choice for ima- various widely available diagnostic methods for imaging
ging of colonic cancer is still a double-contrast barium of primary colorectal cancer are reviewed. The second
enema (DCBE), while rectal cancer is preferably imaged and third sections are on virtual colonoscopy, and PET
with transrectal ultrasound (TRUS) or magnetic reso- and image fusion, respectively. The final part describes
nance imaging (MRI). Both the use and clinical impact the image-guided treatment of colorectal liver meta-
of DCBE and CT in the imaging of rectal cancer is stases. The interaction of endoscopy and radiology in
limited nowadays. New imaging techniques for both primary colorectal cancer, and the imaging of colorectal
colonic and rectal cancer are still evolving: virtual liver metastases are beyond the scope of this review.
colonoscopy (CT- or MRI-based), positron emission
tomography (PET), and the combination of PET and
a cross-sectional technique like CT (image fusion). 2. Imaging of colorectal cancer
Colorectal liver metastases are imaged with US, MRI,
helical or multi-detector CT and PET. MRI, helical or 2.1. Introduction

* Corresponding author. Tel.: +31-71-526-4376; fax: +31-71-524-

Most, if not all, colorectal cancers develop via the
8256. adenoma–carcinoma sequence with an adenoma dwell
E-mail address: [email protected] (M.E.J. Pijl). time, in the majority of patients, of two to three
0959-8049/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0959-8049(02)00052-7
888 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898

decades. The adenoma dwell time for colorectal cancer 2.3. Computer tomography
in patients with flawed mismatch repair genes is far less,
sometimes less than 2 years [1,2]. The malignant poten- The value of CT in the initial work-up of colorectal
tial of an adenomatous polyp can be estimated by: size, cancer is limited. Accuracy rates reported for pre-
presence and length of a stalk, extent of villous change operative staging of colorectal cancer ranges between 48
and degree a cellular atypia and dysplasia [3,4]. Size is and 77% [11,12]. In selected groups with patients suf-
probably the most practical predictor of malignancy. fering from rectal cancer, the accuracy might be up to
The risk of cancer increases from virtually zero for 82% [13]. These disappointing numbers are mainly due
polyps smaller than 5–6 mm to 40–50% for polyps to the inability of CT to determine the depth of tumour
larger than 2 cm. Sessile polyps carry a high risk bowel wall invasion and to differentiate malignant from
for colorectal cancer in over 50%, compared with 10% reactive lymph nodes. The former is caused by the lim-
for pedunculated polyps. An increase in the extent ited intrinsic soft-tissue contrast of CT that prohibits
of villous changes in the polyp carries a higher cancer the visualisation of the various layers of the bowel wall,
risk. and the latter because the detection of malignant lymph
nodes is based on size. In addition, the CT appearance
2.2. Barium enema of colorectal cancer, a discrete soft-tissue mass or focal
wall thickening, is non-specific [14]. Despite these lim-
Both colorectal adenomas and colorectal cancers can itations, CT can be used for treatment planning, detec-
be depicted with a barium enema examination. Three tion of distant metastases and assessment of adjacent
out of four determinants for the estimation of malignant organ involvement.
potential of a polyp can be assessed with a barium
enema: size, presence or absence of a stalk and archi- 2.4. Ultrasound
tecture (extent of villous change). The villous tumours
often have a fine granular or reticular appearance on Transabdominal ultrasound is generally not used for
double-contrast barium enemas (DCBE), because bar- the detection of colonic cancer, although many experi-
ium gets trapped in fissures at the surface of the tumour. enced sonographers have probably stumbled over the
The fourth determinant, being cellular atypia and dys- ‘pseudo-kidney’ sign occasionally. In the detection and
plasia, obviously cannot be assessed. staging of rectal cancer, transrectal ultrasound (TRUS)
The barium enema studies can be performed with is an important tool, since TRUS enables the distinction
either single-contrast or double-contrast techniques. of the various layers of the rectal wall [15–17]. With
The sensitivity of properly performed and interpreted TRUS, rectal tumours can be staged according to the
single-contrast versus double-contrast enemas for clini- T(NM) classification: T1 for tumours confined to the
cally important lesions, polyps over 1 cm and colorectal mucosa or submucosa, T2 for tumours invading the
cancers, are similar. The choice to perform either a sin- proper muscle layer, but confined to the rectal wall
gle-contrast or double-contrast enema is therefore (intact outermost serosal layer), and T3 for tumours
determined by the patient’s ability to cooperate and the penetrating the perirectal fat. The reported tumour sta-
likelihood of colon disease [5,6]. DCBE, though, depicts ging accuracy of TRUS ranges between 67 and 93%
polyps less than 1 cm in size better and more reliably [15,16,18]. The most important staging error is over-
[5,6]. The reported sensitivity of DCBE for polyps over staging of T2 tumours as T3 tumours, caused by spon-
1 cm is 33–100% [7–9] and for colorectal cancer 62– taneous (desmoplastic) or iatrogenic inflammation [19].
100% [2,5,7–9]. The articles by Ott [8] and de Zwart and Contrary to normal (hyperechoic) lymph nodes,
colleagues [9] address the wide range of reported sensi- malignant perirectal lymph nodes tend to be hypoe-
tivities of DCBE for polyps and colorectal cancers, and choic, and therefore can be fairly easy depicted by
its probable causes, various types of study biases. TRUS within the echogenic perirectal fat. The para-iliac
In up to 95% of patients, the entire colon can be and para-aortal lymph node chains, though, remain
examined with DCBE, an obviously important feature, invisible with TRUS. The accuracy of lymph node sta-
when it is used as a screening tool for colorectal cancer ging ranges between 62 and 83% [15,18,20].
and its precursors [7]. The diagnosis of colorectal cancer A drawback of TRUS is the sub-optimal examination
also warrants a careful examination of the entire colon of patients with a stenotic tumour, since the probe can-
since 5% of patients have a synchronous colon cancer, not be placed close enough to the (upper margin of the)
and more than one-third have additional adenomatous tumour.
polyps [4].
DCBE is generally well tolerated by patients and 2.5. Magnetic resonance imaging
complications are uncommon [10]. If a patient is clini-
cally ill or has signs of toxic megacolon, one should MRI is most useful in colorectal cancer located in the
refrain from performing a barium enema [6]. rectosigmoid, because image quality is only minimally
 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898 889

degraded by respiratory motion and peristalsis, in this At this time, a full colon cleansing preparation is
relatively fixed part of the colon. Initially, a body coil or required to achieve acceptable results. Orally adminis-
external surface coils were used to acquire the MRI tered barium contrast products to mark or tag faecal
images, but nowadays endoluminal or phased array residue and suppress the need for bowel cleansing are
coils are used [21–24]. MRI images generated with an under investigation. Some investigators have recom-
endoluminal coil demonstrate very high spatial resolu- mended the routine use of a muscle relaxant to prevent
tion, permitting detailed visualisation of both tumour unwanted colonic collapse and spasm, a problem
and rectal wall but, because of the limited field of view, encountered most commonly in the sigmoid colon.
the mesorectum and surrounding pelvic structures are However, the benefit of spasmolytic agents is con-
difficult to evaluate. When phased array coils are used, troversial. Yee and colleagues [25] compared colonic
the spatial resolution is not as high as with endoluminal distension in 60 patients who were scanned in the supine
coils, but it permits imaging with a large field of view and prone positions, and reported no beneficial effect
and, thus, the assessment of mesorectum and surround- from routine glucagon administration. In our experi-
ing pelvic structures is possible (Fig. 1). The superior ence, adequate distension of the colon and avoidance of
demonstration of rectal wall layers achieved by endo- spasm can be obtained if the patients are allowed to
luminal MRI does not improve staging accuracy insufflate their colon themselves.
(approximately 80%) of rectal cancer compared with A single acquisition of the abdomen and pelvis is
phased array MRI [22]. Beets-Tan and colleagues [21] obtained with the patient holding their breath for 20 s.
published that the clinically more important circumfer- Following the supine scan, the helical CT is routinely
ential resection margin in rectal cancer can be predicted repeated with the patient prone. The use of both supine
very accurately and consistently with phased array and prone helical CT datasets helps differentiate mobile
MRI. As with TRUS, the insertion of an endoluminal stool from fixed pathology such as cancers and polyps,
coil can be impaired when dealing with a stenotic rectal allows more even distension of the colon and improves
tumour. visualisation of segments of colon obscured by intra-
luminal fluid [26,27]. Unlike fibre-optic colonoscopy,
sedation is not required and patients are immediately
3. Virtual colonoscopy: a new tool for colorectal cancer discharged from the CT suite without the need for
screening observation or recovery. The patient’s ability to return
to work or other activities of daily living is important
3.1. Introduction when considering societal costs of colorectal cancer
screening programmes.
Virtual colonoscopy or CT colonography is a pro-
mising new method for assessing the colon. Helical CT
is used to generate high-resolution, two-dimensional
axial images of the abdomen and pelvis. Three-dimen-
sional images of the colon simulating those obtained
with conventional colonoscopy can be reconstructed
from the data obtained. Favourable attributes of virtual
colonoscopy include its safety, high patient acceptance,
and ability to provide a full structural evaluation of the
entire colon. Indications for virtual colonoscopy include
screening for polyps, incomplete or failed colonoscopy
and preoperative assessment of the colon proximal to an
occlusive cancer (defined as a tumour that cannot be
traversed endoscopically).

3.2. Technique

Although a variety of scanning techniques have been

described for virtual colonoscopy, the same basic imag-
ing principles apply: cleansing the patient’s colon using
a standard barium enema or colonoscopy bowel pre-
paration, colonic insufflation with room air and thin- Fig. 1. 63-year-old male with T3 rectal cancer. MRI: axial T2-weigh-
ted turbo spin-echo image. Tumour almost completely fills the rectal
section helical CT of the abdomen and pelvis followed lumen and penetrates the muscular wall laterally (right side). The
by off-line computer manipulation of the CT dataset to shortest distance between tumour and the mesorectal fascia (white
facilitate inspection of the colonic wall. arrows) is 12 mm (black double-headed arrow).
890 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898

3.3. Radiation dose which can have a similar appearance when viewed in
profile using the axial images alone.
Radiation dose is approximately 20% lower than the
typical dose for DCBE [28]. 3.5. CT colonography and polyp detection

3.4. Image display Preliminary results indicated that the accuracy of CT

virtual colonoscopy for polyp detection exceeded bar-
Following image acquisition, the CT data can be ium enema and approached conventional colonoscopy.
viewed using a variety of techniques (Figs. 2 and 3). Recent published studies from the Boston Medical
These include simple evaluation of the axial CT images Center (BMC), the Mayo Clinic and the University of
at lung window settings (Fig. 2a), reformatted two- California in San Francisco (UCSF) reported results on
dimensional, images at cross-sectional and orthogonal relatively large patient populations (Table 1). Fenlon
planes to the long axis of the colon, and three-dimen- and colleagues (BMC) [29] prospectively studied 100
sional, endoluminal (simulating conventional endoscopy patients at high risk for colorectal neoplasia. Virtual
images (Fig. 2b)) and extraluminal images (simulating colonoscopy was performed immediately before con-
barium enema images (Fig. 3b)). Many investigators ventional colonoscopy. The entire colon was visualised
now agree, however, that review of 2D images at lung
window settings is sufficient for detection of potential
abnormalities, and 3D postprocessed views are only
necessary for characterisation of these potential lesions.
The main role of the postprocessed 3D views are to dif-
ferentiate polyps from complex normal colonic folds

Fig. 3. 50-year-old male with partial obstruction. (a) Axial computer

tomography (CT) image showing circumferential mass of the sigmoid
colon (arrows). (b) Volumetric rendering of the colonic mucosa from
CT data providing detail similar to double-contrast barium enema
Fig. 2. (a) 1 cm polyp (arrow) demonstrated on 2D axial computer (DCBE) showing a large apple core tumour (arrow) in the sigmoid
tomography (CT) image displayed at lung window settings. Rectal colon. Note that the colon proximal to this occlusive tumour is well
tube (arrowhead). (b) Virtual colonoscopy image showing the 1 cm distended, and assessed by CT colonography. No synchronous tumour
polyp (crosshair). detected.
 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898 891

Table 1
Polyp detection with CT colonography

Study Patients Polyps Per polyp sensitivity (%) Per patient sensitivity (%)
>5 mm
Subcentimetrea 10 mm Subcentimetrea 10 mm

BMC [29] 100 58 82 91 92 96

Mayo Clinic [26] 180 263 47 75 88 85
UCSF [30] 300 223 80 90 93 100
a
Subcentimetre range resembles 6–9 mm for the BMC study, and 5–9.9 mm for the Mayo Clinic and UCSF study.

by virtual colonoscopy in 87 patients and by conven- plications. It is also acceptable and well tolerated by
tional colonoscopy in 89. In 49 patients, a total of 115 patients. Fenlon and colleagues [29] reported that none
polyps and three colorectal cancers were identified. Vir- of their patients requested that virtual colonoscopy be
tual colonoscopy identified all three cancers, 91% of stopped because of discomfort or pain. The actual ana-
polyps that were 10 mm or more in diameter, 82% of lysis of the colon is performed on the CT data at the
polyps 6–9 mm, and 55% of polyps 5 mm or smaller. physician’s workstation and does not require the patient’s
The authors concluded that in patients at high risk for continued presence. Patients can leave immediately after
colorectal neoplasia, virtual and conventional colono- the procedure, since no sedation is required.
scopy have a similar efficacy for the detection of polyps Adenomatous polyps are common, being present in
that are 6 mm or more in diameter. Fletcher and col- 30–50% of persons over 50 years of age. Most measure
leagues (Mayo Clinic) [26] published a series of 180 less than 10 mm in size and, within this subset, the
patients with 420 colonoscopically-proven polyps. probability of malignancy is extremely low and the
These patients were known to have or were suspected to likelihood of any lesion progressing to malignancy is
have colorectal neoplasms, except for 20 patients extremely small. Many non-neoplastic polyps also
recruited from a surveillance population with a low measure less than 10 mm in size. Because many small
prevalence of large polyps. By using supine and prone polyps have no malignant potential, it may be possible
patient positioning, the sensitivity and specificity for the to target only polyps above a certain size threshold (6
identification of patients with polyps 10 mm or greater mm or even 10 mm) for colon cancer screening and
were 85% (82 of 96 patients) and 93% (78 of 84 colonoscopic removal and by doing so, achieve com-
patients), respectively. The sensitivity and specificity for parable benefits to universal polypectomy with con-
the detection of polyps 5 mm or greater were 88% (114 siderable savings in cost and risk. There is a consensus
of 130 patients) and 72% (36 of 50 patients), respec- among researchers that virtual colonoscopy has the
tively. Yee and colleagues (UCSF) [30] published their ability to separate patients with colons containing only
findings in a series of 300 patients. The sensitivity was subcentimetre insignificant polyps or no polyps from
90% (74 of 82) for the detection of polyps 10 mm or patients with clinically significant polyps. Only the latter
larger, 80% (113 of 141) for polyps 5.0–9.9 mm, and group would require therapeutic colonoscopy for poly-
59% (178 of 301) for polyps smaller than 5 mm. The pectomy. As only 3–10% of average risk persons age 50
sensitivity was 94% (64 of 68) for the detection of his- years or over have a polyp 10 mm or greater in size, a
tologically-proven adenomas 10 mm or larger and 82% test that could reliably and accurately identify this
(72 of 88) for histologically-proven adenomas 5.0–9.9 group would result in a dramatic reduction in the num-
mm. CT colonography accurately detected all eight ber of purely diagnostic colonoscopies performed, free-
colorectal cancers present in this series. ing up endoscopy services for those patients requiring
Proper technique is essential for achieving good therapeutic intervention.
results. Results significantly improve if patients are
scanned in the prone and supine position and adequate 3.7. CT colonography and occlusive cancer
colon distension is achieved. Retained fluid in the colon
can be minimised with improvements in bowel prepar- Fenlon and colleagues [32] recently reported another
ation [26,29,31]. As is true for conventional colono- application of virtual colonoscopy, the preoperative
scopy, virtual colonoscopy has a substantial learning assessment of the colon proximal to an occlusive cancer
curve process. (defined as a tumour that cannot be traversed
endoscopically). In 29 patients with occlusive colorectal
3.6. CT colonography as a screening tool cancers, virtual colonoscopy depicted all 29 occlusive
cancers and demonstrated two additional cancers and 24
CT colonography has been shown to be a safe test; polyps in the proximal colon. Both of the synchronous
none of the studies published so far has reported com- cancers were confirmed intra-operatively and resected.
892 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898

Virtual colonoscopy successfully demonstrated the consecutive patients, Hara and colleagues [38] identified
proximal colon in 26 of 29 patients studied compared 30 (11%) patients with highly important extracolonic
with preoperative barium enema that failed to ade- findings, which resulted in further examinations in 18
quately demonstrate the proximal colon in any patient (7%) patients, including US in 8, CT in 13, and
studied. In 3 patients, retained stools proximal to an intravenous pyelography in 1. 6 patients underwent
occlusive tumour prevented a complete virtual colono- surgery because of incidentally discovered findings
scopic evaluation. Morrin and colleagues [33] had similar (incidental abdominal aortic aneurysms greater than 4
results, and found that CT colonography is superior to cm in 2, asymptomatic renal adenocarcinomas in 2,
barium enema to assess the colon proximal to an pneumothorax and inguinal hernia containing bowel in
occlusive tumour. A total of 97% (87 of 90) of all colonic 1 patient each). Almost half of all patients had abnor-
segments were adequately visualised at CT colonography mal extracolonic findings detected at CT colonography.
in patients with obstructing colorectal lesions (15 patients No other colorectal screening examination has the
were referred after incomplete colonoscopy) compared advantage of detecting potentially life-threatening con-
with 62% (26 of 42) of segments at barium enema. ditions in an age group where those conditions are
prevalent.
3.8. CT colonography and incomplete colonoscopy
3.10. Conclusions
Colonoscopy fails to visualise the caecum in a sig-
nificant number of patients, the completion rate may be Results of virtual colonoscopy published in the recent
80–90% or less. Thiis-Evensen and colleagues [34] pub- literature are extremely encouraging. Software techni-
lished their findings in a total of 241 men and women ques designed to improve the speed, accuracy and
randomly selected from the population register who reproducibility of results are rapidly emerging. The
were offered a colonoscopic screening examination to challenge remains to reproduce these favourable results
detect and remove polyps. The caecum was reached in in clinical practice and to evaluate the use of virtual
only 193 (80%) patients. More than half of the adeno- colonoscopy in a purely screening population. If virtual
mas detected were localised proximal to the sigmoid colonoscopy proves to be an accurate, reliable and cost-
colon and, in nearly half of the adenoma-bearing sub- effective method for detecting polyps and early cancers,
jects examined, the adenoma was proximal to the de- it may dramatically improve participation among the
scending colon. Nicholson and colleagues [35] reported population in screening programmes and play a major
a similar distribution of adenomas in 1131 asympto- role in minimising the impact of colorectal cancer.
matic individuals who underwent full colonoscopy. In
25% of people found to have adenomas, the adenomas
were found only in the proximal colon. This emphasises 4. The role of positron emission tomography in
the importance of a screening method being able to view colorectal cancer
the entire colon.
Virtual colonoscopy excels in the evaluation of the 4.1. Introduction
ascending colon, particularly the caecum due to the
degree of distension achievable and the typical lack of Detection, staging and assessing/predicting response
spasm or muscular hypertrophy which is seen in the to therapy are long-term goals for imaging in colorectal
sigmoid colon. It is therefore a useful complement to an cancer. Anatomical methods have been key in this area
incomplete colonoscopy [36,37]. Morrin and colleagues and are likely to remain very important, but have lim-
[36] prospectively studied 40 patients in whom the cae- itations. Anatomical methods generally detect cancer by
cum could not be reached endoscopically, despite detecting a mass lesion. Tumours must grow to a certain
adequate bowel preparation. CT colonography was volume to be detectable. Furthermore, small tumours
performed within 2 h of incomplete colonoscopy in may not be detected, as they cannot be distinguished
these 40 patients. In addition, 26 patients (65%) under- from normal variants in tissue anatomy. This is true for
went barium enema immediately after CT colono- primary lesions and for the most common initial site of
graphy. CT colonography was better tolerated than metastases, regional lymph nodes. Metastases to regio-
barium enema. CT colonography adequately revealed nal lymph nodes are detected by size criteria in imaging,
96% of all colonic segments; in comparison, barium with nodes larger than 1 cm in diameter viewed as being
enema adequately revealed 91% of all segments. most consistent with malignancy. Anatomical methods
have reported sensitivities for tumour detection of fewer
3.9. Extracolonic findings than 50%. Larger masses are more likely malignant, but
this is not a given fact, as many larger lesions are simply
Since the entire abdomen and pelvis are scanned, the reactive tissue or scars. Detection of masses is also
adjacent abdominal organs can also be assessed. In 264 dependent on comparisons with normal anatomy.
 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898 893

Normal anatomy is markedly distorted by surgery and 4.3. PET and primary colorectal cancer
radiation. Such surgical/postoperative changes can
make the diagnosis of tumour recurrence impossible. PET is being used quite commonly in colorectal can-
Anatomical methods do not predict response to treat- cer management in institutions in which it is available.
ment. Thus, methods to complement anatomical imag- It is a reimbursed technique by Medicare in the USA,
ing are needed. and this has heightened awareness of its utility. PET has
not been explored extensively in the diagnosis of pri-
4.2. Technique mary colorectal cancer. While PET can detect primary
colorectal cancers with a high sensitivity, there are lim-
Positron emission tomography (PET) imaging is a itations which have decreased interest in this approach.
very sensitive quantitative nuclear medicine technique In brief, there is substantial normal FDG activity in the
which detects altered function in tissues, as opposed to gut due to intestinal flora and due to normal uptake of
detecting masses alone. In PET, a positron emitter FDG in muscle, etc. This background tracer uptake makes
labelled substance is injected intravenously and allowed the detection of small lesions challenging and this area
to distribute through the body. After the injection, the has not been explored for screening applications due to
patient is imaged with a PET camera, which looks much concerns of false-positive interpretations. Thus, the
like a CT scanner. The positron is not directly detected apparent low specificity in this setting in limited studies
but, rather, photons released from positron decay are makes evaluation of primary lesions challenging at pre-
detected. Such detection is very efficient due to the use sent, at least without anatomical correlates. Of interest,
of specialised electronics to resolve the time of arrival of is that uptake in hyperplastic polyps to date has been low.
the photons at the camera detectors. A variety of bio- PET has been used to some extent for primary staging
chemical alterations exist in tumours which can be of colorectal cancer. However, PET is not capable of
imaged by PET with the proper choice of tracer, determining the extent or depth of tumour involvement.
including increased glucose metabolism, increased pro- For detecting nodal metastases, PET also has size lim-
tein and DNA synthesis, altered receptor phenotypes itations in detection. A prospective trial by Nabi and
among many others. PET tracer choices are limited by colleagues [39] showed a sensitivity of PET of only 29%
the available positron emitting isotopes, which have for the detection of nodal metastases. This was com-
short half lifes, from 2 to 109 min (Table 2). The PET parable to CT. Thus, PET should not be recommended
isotopes are made in a cyclotron, which must be rela- for routine use in assessing lymph nodes.
tively near the PET camera.
The most commonly used PET radiopharmaceutical 4.4. PET and local tumour recurrence
for colorectal cancer imaging, and for cancer imaging in
general is [18F]fluoro-2-deoxyglucose (FDG). This agent PET has had much better results in detecting distant
is handled much like glucose in the early stages of metastases and tumour recurrence. This is the clinical
tumour glycolytic metabolism. Most notably, it is situation in which PET is commonly used. One of the
transported like glucose into cancer cells and then earliest reports on the use of PET for clinical imaging
phosphorylated by hexokinase to FDG-6-phosphate, a was with FDG in 1982, showing the feasibility of
polar molecule retained in cancer cells. Overexpression detecting distant metastases of colorectal cancer [40].
of facilitative glucose transporters and glycolytic However, the first larger follow-up study compared
enzymes is typical of colorectal cancer. PET images of PET and CT for local recurrence following abdominal
cancers detect FDG-6-phosphate in the cancer cells. perineal resection for rectal cancer. In these two studies,
PET imaging with modern scanners is often more sen- increased FDG uptake was noted in the patients with
sitive and specific than anatomical methods for cancer tumour versus those with scar. Indeed, approximately
detection. Modern PET cameras and FDG allow imag- 5-fold more uptake was seen in recurrent tumour
ing of most of the body in less than 1 h. patients than in those with scarring. The quantitative
ability of PET was useful in these small studies [41,42].
Anatomical methods were unable to make the differ-
Table 2 entiation between benign disease and recurrence.
Radioisotopes for PET
4.5. PET and distant colorectal metastases
Radioisotope Half-life (min)
15
Oxygen 2 A substantial number of manuscripts have been pub-
13
Nitrogen 10 lished in the past several years in which more extended
11
Carbon 20
18
portions of the body were evaluated by PET, using
Fluorine 110
‘whole body methods’ (Fig. 4). These allow evaluation
PET, positron emission tomography. of the chest and abdomen, rather than limited regions of
894 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898

Fig. 4. Whole body [18F] fluoro-2-deoxyglucose (FDG) positron emission tomography (PET) images (left to right: axial, coronal and sagittal view)
showing three metastases (black spots) of colorectal cancer to the liver. Black areas in the renal collecting system and bladder represent excreted
tracer in the GU system. Uptake at upper aspect of middle image is in the brain, which is normal.

the body. These studies have shown that PET is more cancer, PET can be very useful in locating the source of
sensitive than CT for recurrence, PET is more specific recurrence, with over half of the patients having a
than CT and patient management is changed by the source of the rising tumour marker identified [45,46].
availability of the PET data. In assessing treatment response, early studies by
A representative study of Delbeke and colleagues [43] Strauss and colleagues [41] and Ito and colleagues [42]
showed the accuracy of PET to be 92% for extrahepatic showed good accuracy of PET in assessing areas after
metastasis detection versus 71% for CT, with manage- treatment. However, Haberkorn and colleagues [47]
ment changed in 28% of patients. One of the largest cautioned that early after radiation therapy, false-posi-
direct comparative trials of PET and CT for accuracy in tive results could be observed with PET, due to
colorectal cancer, in 105 patients, showed the overall ‘inflammation’. It is generally believed that a delay of
sensitivity of PET to be 87% versus 68% for CT, and the several months after radiation therapy is completed will
sensitivity for liver metastases superior to that of CT (89% reduce the likelihood of false-positive PET scans.
versus 71%). Of note in this study was that mucinous Data available on chemotherapy monitoring indicate
tumours were less well detected than other histologies that declines in tumour glucose metabolism are quite
(sensitivity 58% versus 92%), possibly due to their lower rapid and assessment can be made in a matter of a few
cellularity than other tumours. While CT-angio-porto- weeks after treatment. The precise role of PET in asses-
graphy is a sensitive technique, it has false-positives. PET is sing chemotherapy treatment response is still in evolu-
probably also superior to this method in the liver [43]. tion but, clearly, rapid declines in tumour glucose
Results from eleven studies with a total of 281 metabolism are a desirable effect of treatment and seem,
patients were recently evaluated as part of a meta-ana- from the body of literature available across a variety of
lysis. These showed a sensitivity of 97% for PET and a types of tumours, to be predictive of a favourable
specificity of 76% [44]. PET is superior to CT in all tumour response.
locations, with the possible exception of the lungs,
where metastasis detection is probably comparable [45]. 4.7. Cost-effectiveness of PET

4.6. PET and follow-up PET is a more expensive study at most centres than a
CT scan. Is it cost-effective? Two studies have shown
In the setting of rising carcinoembryogenic antigen that it is, obtaining cost saving predictions of US $220
(CEA) levels after ‘definitive’ treatment for colorectal to over US $3000 per case in which PET is performed
 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898 895

by avoiding unneeded or inappropriate surgical pro- lesions to examine possible liver metastases, which
cedures [45,48]. The precise cost savings depend on the might be surgically treated at the time of initial resection.
patient population. It should be cautioned that these are However, the main role of PET in colorectal cancer
model calculations and may not reflect what happens in has been in detection of recurrence. Similarly, following
practice if the test is applied to patient populations with a rising CEA level, PET is often the first test performed
a lower prevalence of disease. after the CT examination is not informative and may
ultimately be performed before CT. Less data exist
4.8. Image fusion regarding the use of PET for routine surveillance, but in
patients with clinical symptoms suggesting recurrence,
PET is a relatively new imaging method and like other PET should probably be considered. The optimal use
methods, one in which there is ongoing growth and and frequency of imaging tests for detecting recurrent
development of the imaging technology. One of the tumour is controversial. In colorectal cancer where iso-
exciting opportunities in PET at present is the fusion of lated metastasis resection can be useful, a stronger case
anatomical imaging methods and PET into hybrid images can be made for routine PET imaging follow-up at reg-
(Fig. 5). This can be done by software methods (anato- ular intervals than in other, less treatable, conditions.
metabolic fusion) or by dedicated hardware devices, in However, the evidence base to support such conclusions
which PET and CT are performed in the same instru- is not yet firm. In the meta-analysis by Huebner and
ment, a ‘PET-CT’ scanner. These devices will allow us colleagues [44], PET changed management in 29% of
to combine the best in anatomical and functional ima- patients.
ging, although it is not yet proven that they improve
diagnostic accuracy. None the less, our initial experi-
ence with such a device has been very encouraging. 5. Local treatment of colorectal liver metastases

4.9. Conclusions 5.1. Introduction

PET with FDG is an important tool for colorectal Approximately one-half of all colorectal cancer patients
cancer imaging. The role of PET for primary lesion will develop liver metastases. In 25% of cases, liver
detection is not established nor is the data on lymph metastases are present at the time of colorectal cancer
node staging promising. PET is very accurate at detect- diagnosis (synchronous metastases) and a further 25%
ing liver metastases and it could be argued that PET will develop liver metastases within 3 years (metachronous
would be reasonable in patients with larger primary metastases) [49–52]. Without treatment, most patients with

Fig. 5. Coronal reformatted computer tomography (CT) (left), positron emission tomography (PET) (centre) and fusion hybrid images (right) of
multiple colorectal liver metastases. The metastases appear slightly hyperdense in a steatotic liver on the CT image, black on the PET image and
bright on the fused image.
896 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898

liver metastases will die between 6 and 11 months, thermal damage to tissue by either heating to more than
depending on the amount of liver parenchyma involve- 50  C or freezing below 30  C: denaturation of cellular
ment [49]. Some subgroups of patients survive for more proteins, leading to cell membrane rupture and dehy-
than 2 years; 5-year survival without treatment is rare dration, but also coagulation and thrombosis of vessels.
[49]. Unfortunately, chemotherapy and radiation ther- The combined effect leads to tumour destruction.
apy are ineffective treatment methods with no or limited Ablations using laser, microwave and radio-frequency
improvement in median survival rates, surgical resection can be performed in a percutaneous approach as well as
being the only potential curative option [53]. open (via laparotomy). Cryo-ablation is virtually always
Surgical resection of colorectal liver metastases in performed in an open procedure. Correct placement of
patients with liver-only solid tumour metastases, how- the thermal device in or at the tumour is the hallmark of
ever, is only possible in 10–15% of patients due to pre- all methods and is nearly always image-guided (CT,
sence of bilobar disease, unfavourable biology with the transabdominal US or intra-operative US).
presence of more than four metastases, tumour location
preventing the achievement of at least a 1 cm tumour- 5.4.1. Laser ablation
free margin [54]. So, in the majority of patients, other A reproducible thermal injury can be produced with
means of controlling or potentially curing liver metas- neodymium yttrium aluminium garnet (Nd YAG) lasers
tases must be explored [54]. Furthermore, combined [53]. From a single, bare 400-mm laser fibre, light at
approaches in which surgical resection of one or more optical or near-infrared wavelengths will scatter within
segments and ablation in other segments are simulta- tissue and be converted into heat. Lesions up to 2 cm
neously performed, will become increasingly important. can be made. Larger lesions need multiple fibres or the
With continued improvements in technology and use of cooled-tip diffuser fibres. In the Middlesex Hos-
increasing clinical experience, one or more of the pital, London, UK, Lees has reached survival figures for
following minimally invasive techniques may soon inoperable colorectal cancer metastases comparable
challenge surgical resection as the treatment of choice with those for patients with operable metastases using
for patients with limited hepatic tumour involvement laser ablation [53]. A definite advantage of the method
[53]. is its MR compatibility and visualisation of temperature
change and coagulation [53].
5.2. Percutaneous ethanol injection
5.4.2. Microwave ablation
Within neoplastic cells, ethanol causes dehydration In this type of ablation, the basic mechanism of heat
and subsequent coagulation. Within vessels ethanol generation consists of the rotation of water molecules
induces necrosis of endothelium and platelet aggrega- [53]. The rotation follows the alternating electric field
tion, leading to thrombosis and ischaemia [53]. Unlike component of the 2450 MHz microwaves. Ablation
hepatocellular carcinoma, for which ethanol injection is sizes of a single electrode are comparable to laser coag-
very effective, colorectal metastases are firm and have ulation. Experience with this method comes mainly
no surrounding capsule making adequate deposition of from Japan concerning hepatocellular carcinoma abla-
ethanol in these metastases difficult and uncontrolled, tion.
leading to spread into surrounding liver tissue and in
the needle tract [49]. The same holds true for the 5.4.3. Radio-frequency ablation
deposition of boiling saline and acetic acid. The principle of radio-frequency ablation depends on
ionic agitation of molecules leading to frictional heat-
5.3. Hepatic artery embolisation ing. The tissue around the electrode acts as a heat
source. A 450–480 kHz electric current flows from a
The hepatic artery carries out 90% of tumour blood single electrode with multiple tines or cluster electrode,
supply and, tumours being sensitive to ischaemia, it depending on the manufacturer, to ground pads placed
stands to reason to expect embolisation of the hepatic on the patient’s thighs. To prevent charring of tissue,
artery, with or without adding chemotherapeutical which is detrimental to energy deposition in the tissue,
agents, to be effective. In clinical trials, significant several methods are used ranging from saline infusion to
improvement of survival has, however, not been shown cooled tip electrodes. In addition, feedback of the tem-
[55,56]. perature at the tips of the tines and, thus, modulating
output power can diminish the occurrence of charring.
5.4. Thermal ablation The creation of coagulation areas up to 5 cm is possible
[53].
Several thermal ablation methods are in operation Complete ablation rates of 52–71% and 5-year survi-
today: laser, microwave, radio-frequency and cryo- val rates comparable to surgery have been described
ablation. They all rely on the principle of inflicting [49,53,57].
 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898 897

5.4.4. Cryo-ablation 20. Tio TL, Coene PP, van Delden OM, Tytgat GN. Colorectal car-
Freezing tissues to below 30  C will destroy cell cinoma: preoperative TNM classification with endosonography.
Radiology 1991, 179, 165–170.
membranes by the formation of ice crystals and will
21. Beets-Tan RG, Beets GL, Vliegen RF, et al. Accuracy of mag-
subsequently destruct cellular structures [49]. Treatment netic resonance imaging in prediction of tumour-free resection
complications are higher than above-mentioned mod- margin in rectal cancer surgery. Lancet 2001, 357, 497–504.
alities [53], but very much larger lesions (6–8 cm) can be 22. Blomqvist L, Holm T, Rubio C, Hindmarsh T. Rectal tumours—
treated and the ice ball formation can be monitored by MR imaging with endorectal and/or phased-array coils, and his-
intra-operative US. Another drawback of this system is topathological staging on giant sections. A comparative study.
Acta Radiol 1997, 38, 437–444.
the high price of the equipment. Cryo-ablation is the 23. Schnall MD, Furth EE, Rosato EF, Kressel HY. Rectal tumor
oldest of local thermal ablation techniques. stage: correlation of endorectal MR imaging and pathologic
findings. Radiology 1994, 190, 709–714.
24. Brown G, Richards CJ, Newcombe RG, et al. Rectal carcinoma:
References thin-section MR imaging for staging in 28 patients. Radiology
1999, 211, 215–222.
1. Peltomaki P, Aaltonen LA, Sistonen P, et al. Genetic mapping of 25. Yee J, Hung RK, Akerkar GA, Wall SD. The usefulness of glu-
a locus predisposing to human colorectal cancer. Science 1993, cagon hydrochloride for colonic distention in CT colonography.
260, 810–812. AJR Am J Roentgenol 1999, 173, 169–172.
2. Stevenson GW. Colorectal cancer imaging: a challenge for radi- 26. Fletcher JG, Johnson CD, Welch TJ, et al. Optimization of CT
ologists. Radiology 2000, 214, 615–621. colonography technique: prospective trial in 180 patients. Radi-
3. Gore RM. Colorectal cancer. Clinical and pathologic features. ology 2000, 216, 704–711.
Radiol Clin North Am 1997, 35, 403–429. 27. Chen SC, Lu DS, Hecht JR, Kadell BM. CT colonography: value
4. Levine MS, Rubesin SE, Laufer I, Herlinger H. Diagnosis of of scanning in both the supine and prone positions. AJR Am J
colorectal neoplasms at double-contrast barium enema examina- Roentgenol 1999, 172, 595–599.
tion. Radiology 2000, 216, 11–18. 28. Hara AK, Johnson CD, Reed JE, et al. Reducing data size and
5. Gelfand DW. Colorectal cancer. Screening strategies. Radiol Clin radiation dose for CT colonography. AJR Am J Roentgenol 1997,
North Am 1997, 35, 431–438. 168, 1181–1184.
6. Smith C. Colorectal cancer. Radiologic diagnosis. Radiol Clin 29. Fenlon HM, Nunes DP, Schroy III PC, Barish MA, Clarke PD,
North Am 1997, 35, 439–456. Ferrucci JT. A comparison of virtual and conventional colono-
7. Gazelle GS, McMahon PM, Scholz FJ. Screening for colorectal scopy for the detection of colorectal polyps. N Engl J Med 1999,
cancer. Radiology 2000, 215, 327–335. 341, 1496–1503.
8. Ott DJ. Accuracy of double-contrast barium enema in diagnosing 30. Yee J, Akerkar GA, Hung RK, Steinauer-Gebauer AM, Wall
colorectal polyps and cancer. Semin Roentgenol 2000, 35, 333– SD, McQuaid KR. Colorectal neoplasia: performance character-
341. istics of CT colonography for detection in 300 patients. Radiology
9. de Zwart IM, Griffioen G, Shaw MP, Lamers CB, de Roos A. 2001, 219, 685–692.
Barium enema and endoscopy for the detection of colorectal 31. Fletcher JG, Johnson CD, MacCarty RL, et al. CT colono-
neoplasia: sensitivity, specificity, complications and its determi- graphy: overcoming problems of collapse and colonic fluid.
nants. Clin Radiol 2001, 56, 401–409. Radiology 1998, 209, 296.
10. Blakeborough A, Sheridan MB, Chapman AH. Complications of 32. Fenlon HM, McAneny DB, Nunes DP, Clarke PD, Ferrucci JT.
barium enema examinations: a survey of UK consultant radi- Occlusive colon carcinoma: virtual colonoscopy in the preoperative
ologists 1992 to 1994. Clin Radiol 1997, 52, 142–148. evaluation of the proximal colon. Radiology 1999, 210, 423–428.
11. Berlin JW, Gore RM, Yaghmai V, Newmark GM, Miller FH. 33. Morrin MM, Farrell RJ, Raptopoulos V, McGee JB, Bleday R,
Staging of colorectal cancer. Semin Roentgenol 2000, 35, 370–384. Kruskal JB. Role of virtual computed tomographic colonography
12. Horton KM, Abrams RA, Fishman EK. Spiral CT of colon in patients with colorectal cancers and obstructing colorectal
cancer: imaging features and role in management. Radiographics lesions. Dis Colon Rectum 2000, 43, 303–311.
2000, 20, 419–430. 34. Thiis-Evensen E, Hoff GS, Sauar J, Majak BM, Vatn MH. Flex-
13. Chiesura-Corona M, Muzzio PC, Giust G, Zuliani M, Pucciarelli ible sigmoidoscopy or colonoscopy as a screening modality for
S, Toppan P. Rectal cancer: CT local staging with histopatholo- colorectal adenomas in older age groups? Findings in a cohort of
gic correlation. Abdom Imaging 2001, 26, 134–138. the normal population aged 63–72 years. Gut 1999, 45, 834–839.
14. Thoeni RF. Colorectal cancer. Radiologic staging. Radiol Clin 35. Nicholson FB, Korman MG, Stern AI, Hansky J. Distribution of
North Am 1997, 35, 457–485. colorectal adenomas: implications for bowel cancer screening.
15. Heriot AG, Grundy A, Kumar D. Preoperative staging of rectal Med J Aust 2000, 172, 428–430.
carcinoma. Br J Surg 1999, 86, 17–28. 36. Morrin MM, Kruskal JB, Farrell RJ, Goldberg SN, McGee JB,
16. Rifkin MD, Ehrlich SM, Marks G. Staging of rectal carcinoma: Raptopoulos V. Endoluminal CT colonography after an incom-
prospective comparison of endorectal US and CT. Radiology plete endoscopic colonoscopy. AJR Am J Roentgeno 1999, 172,
1989, 170, 319–322. 913–918.
17. Yamashita Y, Machi J, Shirouzu K, Morotomi T, Isomoto H, 37. Macari M, Berman P, Dicker M, Milano A, Megibow AJ. Use-
Kakegawa T. Evaluation of endorectal ultrasound for the fulness of CT colonography in patients with incomplete colono-
assessment of wall invasion of rectal cancer. Report of a case. Dis scopy. AJR Am J Roentgenol 1999, 173, 561–564.
Colon Rectum 1988, 31, 617–623. 38. Hara AK, Johnson CD, MacCarty RL, Welch TJ. Incidental
18. Beynon J. An evaluation of the role of rectal endosonography in extracolonic findings at CT colonography. Radiology 2000, 215,
rectal cancer. Ann R Coll Surg Engl 1989, 71, 131–139. 353–357.
19. Hulsmans FJ, Tio TL, Fockens P, Bosma A, Tytgat GN. 39. Abdel-Nabi H, Doerr RJ, Lamonica DM, et al. Staging of
Assessment of tumor infiltration depth in rectal cancer with primary colorectal carcinomas with fluorine-18 fluorodeoxy-
transrectal sonography: caution is necessary. Radiology 1994, glucose whole-body PET: correlation with histopathologic and
190, 715–720. CT findings. Radiology 1998, 206, 755–760.
898 M.E.J. Pijl et al. / European Journal of Cancer 38 (2002) 887–898

40. Yonekura Y, Benua RS, Brill AB, et al. Increased accumulation 48. Gambhir SS, Valk P, Shepherd J, Hoh C, Allen M, Phelps ME.
of 2-deoxy-2-[18F]fluoro-D-glucose in liver metastases from colon Cost effective analysis modeling of the role of FDG PET in the
carcinoma. J Nucl Med 1982, 23, 1133–1137. management of patients with recurrent colorectal cancer. J Nucl
41. Strauss LG, Clorius JH, Schlag P, et al. Recurrence of colorectal Med 1997, 38(P), 90.
tumors: PET evaluation. Radiology 1989, 170, 329–332. 49. Rovers JP. Photodynamic Therapy for Colorectal Liver Metas-
42. Ito K, Kato T, Tadokoro M, et al. Recurrent rectal cancer and tases: A Preclinical Study. Thesis, Leiden University Medical
scar: differentiation with PET and MR imaging. Radiology 1992, Center, Leiden, The Netherlands, 2001, 7–8.
182, 549–552. 50. Asbun HJ, Hughes KS. Management of recurrent and metastatic
43. Delbeke D, Vitola JV, Sandler MP, et al. Staging recurrent colorectal carcinoma. Surg Clin North Am 1993, 73, 145–166.
metastatic colorectal carcinoma with PET. J Nucl Med 1997, 38, 51. Ballantyne GH, Quin J. Surgical treatment of liver metastases in
1196–1201. patients with colorectal cancer. Cancer 1993, 71, 4252–4266.
44. Huebner RH, Park KC, Shepherd JE, et al. A meta-analysis of 52. Foster JH. Surgical treatment of metastatic liver tumors.
the literature for whole-body FDG PET detection of recurrent Hepatogastroenterology 1990, 37, 182–187.
colorectal cancer. J Nucl Med 2000, 41, 1177–1189. 53. Dodd III GD, Soulen MC, Kane RA, et al. Minimally invasive
45. Valk PE, Abella-Columna E, Haseman MK, et al. Whole- treatment of malignant hepatic tumors: at the threshold of a
body PET imaging with [18F]fluorodeoxyglucose in manage- major breakthrough. Radiographics 2000, 20, 9–27.
ment of recurrent colorectal cancer. Arch Surg 1999, 134, 503– 54. Curley SA, Izzo F, Delrio P, et al. Radiofrequency ablation of
511. unresectable primary and metastatic hepatic malignancies: results
46. Flanagan FL, Dehdashti F, Ogunbiyi OA, Kodner IJ, Siegel BA. in 123 patients. Ann Surg 1999, 230, 1–8.
Utility of FDG-PET for investigating unexplained plasma CEA 55. Bengmark S, Jeppsson B. Status of ischemic therapy for hepatic
elevation in patients with colorectal cancer. Ann Surg 1998, 227, tumors. Surg Clin North Am 1989, 69, 411–418.
319–323. 56. Pentecost MJ. Transcatheter treatment of hepatic metastases.
47. Haberkorn U, Strauss LG, Dimitrakopoulou A, et al. PET stud- AJR Am J Roentgenol 1993, 160, 1171–1175.
ies of fluorodeoxyglucose metabolism in patients with recurrent 57. Solbiati L, Goldberg SN, Ierace T, et al. Hepatic metastases:
colorectal tumors receiving radiotherapy. J Nucl Med 1991, 32, percutaneous radio-frequency ablation with cooled-tip electrodes.
1485–1490. Radiology 1997, 205, 367–373.