This study included 111 client-owned dogs that had undergone CT imaging and surgery to treat thoracolumbar intervertebral disc extrusion(s) and secondary myelopathy at the Virginia-Maryland Regional College of Veterinary Medicine between May 2005 and September 2006. Animals were included if 2D and MPR CT image file sets were available in the hospital's digital image archive and if a compressive myelopathy secondary to extruded disc material within the vertebral canal had been confirmed at surgery. Dogs who had myelographic contrast injections prior to CT image acquisition were excluded. Dogs who had previous thoracolumbar surgery were also excluded.

The first author reviewed medical records and recorded clinical data for each dog. Information recorded from the medical records included the patient name, signalment, body weight on the date of surgery, and the complete surgical report from that visit. Data recorded from the surgical report included the surgical diagnosis, surgical site (disc space), and the side (left, right, bilateral, or mixed) on which disc material was found.

All 2D transverse CT images were acquired using the same single detector spiral CT scanner (Picker PQ5000; Universal Medical Systems, USA). Dogs were placed under general anesthesia and positioned in dorsal recumbency. The standard scanning protocol consisted of transverse slices from mid-T10 to mid-L3 [25]. Additional slices were obtained if requested by the primary care clinician or duty radiologist. Slice thickness settings ranged from 2~3 mm, with a 1 mm slice overlap. Two-dimensional CT image sets for each dog were converted into Digital Imaging and Communications in Medicine (DICOM) format and transferred via Ethernet to a Picture Archiving and Communication System (PACS) (RapidStudy; Eklin Medical Systems, USA). Immediately after scanning, studies were also transferred via Ethernet to a CT workstation (Voxel Q Visualization Station; Picker/Philips Medical Systems, USA). Multi-planar reformatted CT images for each dog were created by the on-duty radiologist, using the workstation's image analysis software. Oblique transverse MPR images were generated using the oblique MPR tool, with the slice angle oriented perpendicular to the long axis of the vertebral canal (Fig. 1). Oblique sagittal MPR images were generated using the same tool, with the slice angle oriented parallel to the long axis of the vertebral canal (Fig. 2). Curved dorsal MPR images were created using the curved MPR tool, with a hand-traced line of cut along the dorsal margins of vertebrae (Fig. 3). Each MPR image was saved as a screen capture and the set of saved images was transferred to the PACS via Ethernet. The MPR image set was stored in the PACS as a separate file for each dog, with the identifier "screen save" included in the file name.

Two board-certified veterinary surgeons and one board-certified veterinary neurologist independently reviewed the digital CT image sets for each included dog. Readers were unaware of clinical and surgical findings at the time of review. Readers retrieved CT image sets from the PACS using an in-hospital network and reviewed images at a diagnostic workstation using standard DICOM viewing software (e-Film; Merge Healthcare, USA). Each reader reviewed 2D CT image sets first, with the list of cases arranged in random order. After they had completed review of 2D CT images, readers were then given a re-randomized list of cases and asked to review MPR CT image sets. Readers recorded their opinions using questionnaires. One questionnaire was created for each of the 2D CT file sets and a separate questionnaire was created for each of the MPR file sets. In the 2D CT questionnaire, readers were asked to record their localization of the extruded disc material and their prediction for which side was affected (left, right, bilateral, or mixed). The questionnaires allowed the participants to write in their lesion localization free hand. Participants were instructed to be as descriptive as possible concerning the site and extent of the extruded disc material. Readers used a numerical rating system (1~10) to indicate their certainty for each diagnosis. For the MPR CT questionnaires, readers were asked to record the same data as above. They were also asked to record which MPR views were available, using the following designations: oblique transverse, curved dorsal planar, sagittal planar, or other. For the available views, readers were asked to choose which they found to be the most helpful in making the diagnosis. An area for general comments was also provided at the end of the questionnaire so that readers could note any other factors that they felt affected their decision-making.

The first author compared CT localizations recorded on reader questionnaires with surgical localizations recorded in medical records. Lesions found in the region extending from the caudal 1/4 of the cranial vertebra to the cranial 1/4 of the caudal vertebra were defined as being located over an intervertebral disc space (Fig. 4). Lesions found in the middle 1/2 of the vertebral body were defined as being located over a vertebra. The site of each lesion was assigned to an arbitrary numbering system that was used for statistical analyses (Fig. 5). Starting at the T8-T9 disc space and ending with the L7-S1 disc space, a number from 1~13 was assigned to each adjacent disc space. Lesions located over vertebral bodies were assigned the cranial disc space's number and a 0.5 value. A reader's localization for CT lesion site was defined as "correct" if at least one of the sites identified by the reader was also identified in the surgical report. A reader's CT localization for lesion extent of involvement was defined as "correct" if it agreed with the cranio-caudal extent of the lesion described in the surgical report. No distinction was made if the reader identified more or fewer lesion sites than those reported at surgery. The reader's CT localization for the affected side (left, right, bilateral, mixed) was defined as "correct" if it agreed with the affected side in the surgical report.

A statistician selected and performed all analyses using statistical analysis software (SAS version 9.1.3; SAS, USA). Reader diagnostic accuracy ("correct" or true positive fraction) estimates for 2D CT images were compared to diagnostic accuracy estimates for MPR CT images using Generalized Estimating Equations. Mean certainty scores for 2D and MPR CT images were compared using a paired Student's t-test, after verifying that the paired differences were normally distributed. Only the "correct" reader responses were used for the certainty comparisons. Kappa statistics were used to assess agreement among readers for 2D versus MPR CT images. The frequencies with which MPR views were chosen as "most useful" were compared using Exact Chi-square statistics. For all analyses, a value of p ≤ 0.05 was considered significant.

The patient population consisted of 60 males, of which 43 were castrated and 17 were intact. There were 51 females within the group, and 45 of those patients were spayed while six were intact. The Dachshund was the most common breed in the study population, with 63 individuals included. Mixed breed dogs were the second most common, with 16 individuals. Other breeds included the Beagle (8), Pekingnese (3), Shih Tzu (3), Basset hound (2), Cockapoo (2), Jack Russel Terrier (2), Lhasa Apso (2), Poodle (2), Cocker Spaniel (1), Bichon Frise (1), Papillon (1), Peekapoo (1), Rat Terrier (1), Tibetan Spaniel (1), and Wheaton Terrier (1). There was also one German Shepherd Dog that met the inclusion criteria. Patient ages ranged from one year and two months to 11 years and eight months. The average age of the animals included in the study was five years three months, and the median age was six years seven months. The average body weight at the time of surgery was 9.1 kg, with a range of 2.7 kg to 36.8 kg.

A total of 137 disc extrusion sites were confirmed at surgery in these 111 dogs. In all cases, a hemi-laminectomy was performed to decompress the spinal cord. The vertebral column was approached from the left in 61 cases, from the right in 47 cases, and a bilateral approach was made in three cases. The most commonly affected site in our study population was T12-T13 (40/137, 30%). The thoracolumbar junction was the second most commonly affected site (31/137, 23%). Other intervertebral disc spaces that were frequently encountered in this study were T11-T12 (20/137, 15%), L1-L2 (13/137, 9%), L2-L3 (12/137, 9%). Less frequently affected sites included L3-L4 (6/137, 4%), L5-L6 (4/137, 3%), L4-L5 (3/137, 2%), and L6-L7 (2/137, 1%). Two disc extrusions were present exclusively over the vertebral body. One of these lesions was at the L2 vertebrae and the other was at L4.

With 2D CT, the total number of true positive localizations were 102 for reader 1, 104 for reader 2, and 99 for reader 3. With MPR CT, the true positive localizations were 100 for reader 1, 105 for reader 2, and 106 for reader 3. Diagnostic accuracy for identifying the correct lesion site was significantly greater with MPR vs. 2D CT images in one reader (Table 1). A significant increase in mean diagnostic certainty scores was seen in two of the readers for MPR vs. 2D CT images (Fig. 5). Mean diagnostic certainty scores for correct diagnoses were increased from 3~11% with MPR CT images. For all other comparisons, no significant differences were identified.

A trend for increased inter-observer agreement was seen for MPR vs. 2D CT images (Table 2). However the increase was not statistically significant. Agreement among readers was highest for the 'affected site' and lowest for the 'number affected'. The change in agreement between 2D and MPR images differed from zero for all analyses (site, side, number affected) among all three readers (p ≤ 0.05).

Oblique transverse and curved dorsal MPR views were rated most helpful with the highest frequency (Table 3). Two readers preferred the curved dorsal MPR images by a large margin compared to other views. The third reader chose the oblique transverse MPR view more often than any other. Oblique sagittal MPR views were infrequently chosen as most helpful by any of the readers.

One of the most commonly recorded limiting factors was a difficulty in visually distinguishing between hemorrhage and extruded disc material. Another factor often mentioned was the presence of an abnormal rib number. Other difficulties recorded by readers included: distinguishing chronic disc protrusion versus extrusion, subjectively interpreting the severity of spinal cord compression when multiple non-contiguous sites were suspected, and making accurate localizations when there were abnormal numbers of lumbar vertebrae. Difficulty identifying the right/left radiographic marker in MPR images was also mentioned. Most readers commented that more time was required to interpret 2D CT images versus MPR images, however interpretation times were not recorded.