Combined PET/CT colonography: is this the way forward?

Short abstract

Combined whole body positron emission tomography/computed tomography (PET/CT) examination may be beneficial for staging in patients with colorectal cancer

“The Best of Both Worlds” (Star Trek‐The Next Generation, season 3, episode 26, stardate 43989.1)

Computed tomography (CT) colonography is a recently introduced technique which is being investigated for several indications. Its role as a screening tool for polyp detection is still controversial.¹,²,³,⁴,⁵,⁶,⁷ Most studies show that the method has a sensitivity of >90% in detecting colorectal polyps of 10 mm or more in size. However, the influence of the scanner or visualisation hardware and software is not clear.⁸,⁹ Additionally, the learning curve for image interpretation is an important quality and cost factor for CT based colonography.¹ Despite these concerns, the use of CT or magnetic resonance imaging (MRI) based colonography in patients with incomplete colonoscopy is becoming a more and more accepted examination method in experienced clinical centres.¹⁰,¹¹ The major downside of sectional radiological imaging such as CT and MRI is the lack of specific functional data. The only functional information in CT and standard MR imaging is contrast media uptake, which is a rather unspecific feature. On the other hand, functional imaging methods such as [¹⁸F]‐fluoro‐2‐deoxy‐D‐glucose‐positron emission tomography (FDG‐PET) are particularly accurate in staging primary and recurrent colorectal cancer, but suffer from inferior anatomical resolution.¹²,¹³,¹⁴ Consequently, it appears very appealing to integrate FDG‐PET imaging into a high resolution multislice CT examination to have the best of both worlds in one comprehensive data set.¹⁵,¹⁶,¹⁷,¹⁸

In this issue of Gut, Veit and colleagues¹⁹ present a feasibility study applying a whole body PET/CT protocol with additional preparation and distension of the colon, resulting in a comprehensive whole body PET/CT colonography examination (see page 68). Data acquisition was performed using a dual slice CT scanner with an integrated PET system. Studying 14 patients with suspected colorectal cancer, one additional colonic lesion in a patient with incomplete colonoscopy was detected. Lymph node staging proved to be correct in nine out of 11 patients. PET/CT identified increased glucose metabolism, suggesting malignancy in one patient where histopathology showed high grade intraepithelial dysplasia without cancerous growth. Moreover, six additional tumour sites (five of them previously unknown) such as liver metastases, breast cancer, hepatocellular carcinoma, pulmonary metastases, and thyroid carcinoma were identified. Based on this highly selected patient group, the authors conclude that combined PET/CT examination may be beneficial for patients with incomplete colonoscopy.

Even if this comprehensive and expensive combination of different imaging modalities is not suited as a screening tool for polyp detection, the idea of integrating different imaging methods into one comprehensive data representation is very appealing. It is cumbersome, inefficient, and fundamentally difficult to compare PET and CT scans just by mental fusion. Consequently, computer scientists have investigated various approaches towards the automatic or semi‐automatic registration of image data sets. These may be classified as either rigid or non‐rigid registration methods. Rigid registration essentially moves the two different three dimensional data sets using translations and rotations in space to find an optimal match. This method is effective for the registration of rigid anatomical regions, such as bones or the skull. But it may fail for non‐rigid anatomical regions, such as the thorax, abdomen, or pelvis, if the patient breathes or is positioned differently within the two imaging devices or if the bowels move during the two examinations. One way to overcome this problem is dual scanning, which means having two different scanning modalities, such as PET and CT, combined into one major apparatus, as described in by Veit and colleagues.¹⁹ Imaging can then be done in a very time efficient manner and bowel movements can be reduced during a 30 minute period using spasmolytic drugs.

The major drawbacks of this solution are the high investment cost as well as the limitation on certain combinations of modalities. There are enormous technical challenges to integrating, for example, an MRI scanner with a PET and CT scanner because of the magnetic effects. Non‐rigid registration can somewhat overcome these problems by registering the surfaces of organs and structures within two data sets.²⁰ This approach can even be used to register a standard anatomical three dimensional atlas to an individual patient by detecting and registering anatomical similarities.²¹ This methodology has already been used in image guided neurosurgical procedures where brain shift is compensated for in order to register intraoperative imaging with preoperative MR or other image data.²² Even data registration between a prone and supine acquired CT colonography has been successfully performed using a non‐rigid approach.²³ Using this robust algorithm, a dual fly‐through of the colon, presenting synchronised prone and supine scans, is feasible.

Digital integration of nearly every imaging modality in a radiological department offers the perfect foundation for practically every type of data fusion. MRI, CT, as well as ultrasound and flat panel radiography represent the most frequently used radiological tools providing primary digital source data. Theoretically, registration and comprehensive integration of these digital data into a connected three dimensional representation of the human body should be possible.²⁴,²⁵,²⁶ In feasibility studies, even different sectional imaging data such as intraoperative laparoscopic ultrasound and three dimensional CT has been fused successfully²⁷ using a standard laptop computer. Using a contour mapping framework, the fusion of two dimensional projection imaging such as fluoroscopy and three dimensional CT data can be achieved.²⁸,²⁹ Most of the above mentioned image fusion approaches should be considered as work in progress. However, the majority have the potential to be integrated into a comprehensive imaging framework in the near future.

According to the 19th century gestalt psychologist Wolfgang Metzger,³⁰ the sum total is considered to be more than just the sum of its individual components. This can be adapted to radiological image data where the comprehensive mutual image information could increase diagnostic sensitivity and specificity. The present multimodality multisession diagnostic workup could be optimised by multiple scanning techniques, as described by Veit and colleagues,¹⁹ as well as by advanced software approaches which combine multimodality acquisition into one comprehensive three dimensional data set. Probably both methodologies will find their applications. Radiologists and computer scientists will continue to focus on this challenging subject and, to conclude with another Star Trek phrase, “to boldly go where no one has gone before”.