Background Revision of total knee arthroplasty (TKA) requires preoperative assessment to identify the causes of failure. Multidetector computerised tomography (MDCT) is a commonly used imaging technique, but is sensitive to certain artifacts, such as metal implants, limiting its use. Cone-beam CT (CBCT) is a new technique dedicated to musculoskeletal imaging that is less sensitive to artifacts and could be utilised in knee implantation surgery. CBCT has not yet been validated for this indication, and we therefore undertook a retrospective assessment of MDCT versus CBCT, comparing: 1) image quality; 2) reproducibility of angle measurements; 3) effectiveness in screening for periprosthetic radiolucency and implant loosening; and 4) radiation dose. Hypothesis This study hypothesised that CBCT provides better image quality, angle measurement reproducibility, and screening for radiolucency and implant loosening at lower doses of radiation than MDCT. Patients and method Between October 2017 and March 2018, 28 patients, with a mean age of 61 ± 11.6 years [range, 45–85 years] underwent both MDCT and CBCT for pain following TKA. Two radiologists performed angle measurements on both devices: patellofemoral tilt (PFT), rotation angle of the femoral component (RAFC) and rotation angle of the tibial component (RATC). They also screened for pathological radiolucency and/or implant loosening, and assessed image quality at the various bone/implant interfaces. The mean CT dose index per examination was recorded. Results Intraclass correlation coefficients for angles and radiolucency screening on MDCT and on CBCT were respectively good (0.73) and excellent (0.82) for PFT, borderline (0.28) and moderate (0.44) for RAFC, excellent (0.82) and excellent (0.96) for RATC, and moderate (0.45) and excellent (0.84) for radiolucency screening. The inter-observer kappa correlation coefficients for diagnosis of implant loosening and image quality assessment for MDCT and CBCT were respectively moderate (0.45) and excellent (0.93) for tibial loosening and low (0.19) and borderline (0.38) for femoral loosening. The mean image quality at the various interfaces for MDCT and CBCT was respectively 2.2/3 and 2.75/3 at the tibia/tibial implant interface, 1/3 and 2.3/3 at the trochlear region/femoral implant interface, 0.9/3 and 2/3 at the femoral condyle/femoral implant interface, and 1.25/3 and 2.1/3 at the patella/patellar medallion interface. The mean CT dose index was significantly lower, by a factor of 1.24, on CBCT (4.138 mGy) than MDCT (5.125 mGy) (p < 00396). Conclusion The results of the present study revealed added value for CBCT in the etiological work-up for pain following a TKA. It was reliable and reproducible for the rotation measurement and diagnosis of implant loosening, due to enhanced image quality despite a lower radiation dose than conventional MDCT.