Objective The aim of this work is to evaluate the diagnostic accuracy of 0.3T sectoral MR imaging, compared with arthroscopy, for meniscal, cruciate ligaments and chondral knee lesions. Materials and Methods We conducted a retrospective study analyzing all the consecutive knees subjected to arthroscopy at our institution between January 2014 and June 2017 and preceded within 3 months by knee MR examination at our institution with 0.3 T equipment. Patients with history of a new trauma in the time interval between MR exam and arthroscopy were excluded from the study. Two independent experienced radiologists evaluated in double blind the MR findings of menisci, cruciate ligaments and articular cartilage. Both radiological findings were independently compared with those of the arthroscopic report considered as gold standard. For each of the examined targets we calculated the following parameters: sensitivity, specificity, accuracy, positive and negative predictive value; interobserver concordance statistically calculated using Cohen's Kappa test. Results 214 knees (95R/119L) of 214 patients (143M/71F) aged from 18 to 72 years (mean 44) were included and analyzed. We found a good diagnostic accuracy of the low field MR in identifying the injuries of the menisci (93%) and the crossed ligaments (96%), but a lower accuracy for the articular cartilage (85%). Sensitivity resulted 90% for menisci, 73% for ligaments and 58% for cartilage. Specificity was 91% for menisci, 97% for ligaments and 92% for cartilage. Inter-observer concordance resulted to be excellent for cruciate ligaments (K of Cohen's test = 0.832), good (K = 0.768) for menisci, modest to moderate for articular cartilage (K from 0.236 to 0.389) with worse concordance for tibial cartilage. Conclusions Low-field MR sectoral device with dedicated joint equipment confirms its diagnostic reliability for the evaluation of meniscal and cruciate ligaments lesions but is weak in evaluating low grade chondral lesions.