To examine the performance of film compared to DICOM (Digital Imaging and Communications in Medicine) images in the detection and volume appreciation of periprosthetic osteolysis around total knee replacements.

Simulated osteolytic lesions were created around 3 cadaveric total knee replacements and fluoroscopic-assisted radiography as well as Computed Tomography derived imaging taken. Arthoplasty surgeons then reviewed the hard images (AP, Lateral, Paired Obliques and Computed Tomography) on 2 separate occasions regarding the presence and size of lesions. With a minimum of 2 months since the last assessment, DICOM images taken from the same knees were then assessed by the same 3 arthroplasty surgeons in the same manner using Syngo™ Pictured Archive Communication System on hospital computer monitors.

Area under the ROC for lesions detection and kappa statistic for volume appreciation derived from the DICOM assessments were not superior with statistical significance to film assessments. Combinations of imaging that incorporated Paired Oblique views had superior performance in both hard-copy and DICOM imaging.

Digital imaging in this study has not clearly demonstrated superiority to film images for detection and volume appreciation of periprosthetic osteolysis around total knee replacements. The value of the addition of the oblique view to routine assessment is again demonstrated.

Osteolysis commonly causes total knee replacement (TKR) failure, often associated with asymptomatic large defects. Detection and size estimation of lytic defects is important for the indications and planning of revision surgery. Our study compares the utility of fluoroscopic-guided plain X-rays and computed topography (CT) in osteolysis detection and volume appreciation.

Three cadaveric specimens were imaged at baseline and following the creation of reamed defects (small, medium and large approximately = 1, 5 & 10 cm3 volume respectively) in the tibia and femur with TKR component implantation at each timepoint. Imaging was with fluoroscopic-guided plain X-rays (Anteroposterior & Lateral [APL], Paired Oblique [OBL]) as well as rapid-acquisition spiral Computed Topography [CT] with a beam-hardening artefact removal algorithm.

Three arthroplasty surgeons estimated the size of the lesion, if present, and confidence (none=0, fair=1, excellent=2) in their assessment on randomly presented images. Each surgeon performed two assessments of each image one month apart.

The accuracy of detecting lesions was determined using the area under the receiver-operating curve (AU-ROC) obtained from a logistic regression with adjustment for assessment sequence, observer, knee and bone. Volume appreciation and assessor confidence were determined using Kappa and the mean average of confidence scores respectively.

The AU-ROC using combinations of either APL/OBL/CT (0.83) or OBL/CT (0.83) resulted in superior detection of lesions (p<0.05) compared to APL (0.75) or OBL alone (0.77). Correct volume appreciation was highest with APL/OBL/CT (kappa=0.52), followed by APL/OBL (0.51) and was superior (p<0.05) to APL (0.29) or CT alone (0.31). Small and medium defects were more often missed than large with all modalities (20.3 vs. 39.7 %). Femoral defects were missed more often than tibial defects (40% vs. 28.7%) and small lesions missed more with CT (50%) versus APL (48%) and Oblique (40%). CT missed 19% of large sized defects, attributed mostly to femoral (29.1%) rather than tibial defects (8.3%)

Greater confidence was derived from use of CT (1.29) and APL (1.19) [Interquartile range (IQR) 1,2] when compared to OBL (.98, IQR 1,1) [p<.01]. Also, there was greater confidence regarding judgement of tibial defects (1.25, IQR 1,2) compared with femoral defects (1.05, IQR 1,1) [p<.01].

Combining all imaging modalities was synergistic and the most sensitive and specific means of defect detection and volume appreciation. CT provided more confidence, superior detection and volume appreciation when used in combination with APL/OBL versus APL/OBL alone. There is also additional value when APL is combined with OBL.