It is still uncertain to determine the exact diagnostic accuracy of MRI for medial meniscus (MM) tear, lateral meniscus (LM) tear and MM posterior root tear (MMPRT) at different magnetic field intensities (MFIs), different sequences and different publication dates. This study aimed to identify the diagnostic performance of MRI for different meniscus lesions at different MFIs, different sequences and different publication dates, and also to compare it with physical examination. PubMed, Embase, Ovid database, Biosis Previews, Cochrane library, Web of Sciences and manual searching were performed from 1 January 2000 to 31 December 2021. Prospective studies of meniscus injuries examined by physical examination, MRI and arthroscopy were included. Thirteen studies with 1583 meniscal tears were included. The pooled sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR-), diagnostic odds ratio (DOR) and area under curve (AUC) were 87% (95% confidence interval (CI), 86-89%), 82% (80-83%), 7.44 (5.11- 10.75), 0.18 (0.14-0.23), 45.95 (26,68-79.15) and 0.93, respectively. The pooled specificity between MM and LM (p = 0.015), the pooled sensitivity and LR- between MM and MMPR (p = 0.031), different MRI sequences (p = 0.035, p = 0.027), and the accuracy of less than 1.5T and 3.0T (p = 0.04), 1.5T and 3.0T (p = 0.035) were statistically different. There was no publication bias (p = 0.54). MRI performed well in the diagnosis of MM tear, LM tear and MMPRT, and the diagnostic performance of physical examination is similar to MRI. The diagnostic accuracy of 3.0T is the highest, and the -weighted imaging (SWI) sequence may be beneficial for diagnosing meniscus tear. However, there are not enough evidence to prove that recent studies are significantly better than previous ones.