Magnetic Resonance Imaging (MRI) of the pathological knee is sought in order to improve sensitivity and specificity of a diagnosis, thus aiding in decision-making regarding surgery. Arthroscopic findings, however, frequently differ from what was previously observed. Existing studies offer little in regards of knee pathology due to their small sample sizes. Our objective was to compare the diagnostic accuracy in MRI reading of the pathological knee for both radiologists and orthopedic surgeons.

We randomly selected 80 patients from all those submitted to knee arthroscopy between 2013 and 2015 in our institution who had previously been examined clinically and through MRI. Three radiologists and three orthopedic surgeons were invited to integrate their respective groups in MRI reading. Only clinical information was previously available at them, neither revealing patient's identity nor any information achieved at the arthroscopic procedure. Group observations regarding injuries seen at any compartments’ chondral surface, menisci, cruciate ligaments and/or the presence of synovitis were collected. Whenever not in concordance, the in-group majority defined the MRI finding. Acquired data was later compared with arthroscopy. Statistical analysis was ran through IBM SPSS Statistics version 20.0.0, considering a confidence interval (CI) of 95% with statistical significance for a p-value under 0.05.

Mean patients’ age at the time of surgery was 41.57 [CI 95% –38.77; 44.36] years old. Women composed 32 (40%) of cases. The time-lapse between MRI and knee arthroscopy was 6.34 [CI 95%: 5.22; 7.47] months. The left side was affected in 48 (60%) of cases. Radiologist findings achieved statistical significance regarding injuries of the chondral surface of the lateral femorotibial compartment (lateral condyle: p<0.001; lateral plateau: p=0.044), anterior cruciate ligament (p<0.001) and medial meniscus (p<0.001). Orthopedic surgeons overread injuries of the chondral surface for any compartment, although their findings also achieved statistical significance in injuries of anterior cruciate ligament (p=0.003) and medial meniscus (p<0.001), with an improved reading over injuries of the lateral meniscus (p=0.004). When classifying menisci injuries, only orthopedic surgeons discriminated them successfully (p=0.003; p=0.014, respectively). For most of the observations, orthopedic surgeons achieved higher sensitivity while radiologists found higher specificity.

MRI seems to offer safe readings in regards of ACL or medial meniscus injuries, regardless the specialty of the observer. Other structures are not read in such a homogeneous way, possibly compromising the surgery later on. The chondral surface of the medial compartment, in particular, is scarcely well read.

Specialties’ MRI interpreting skills differ considerably for any other than the medial meniscus and ACL. Gathering both radiologists and orthopedic surgeons would possibly increase both sensitivity and specificity of the exams.