The initial online search performed on the 12th of May 2024 yielded 1884 articles excluding duplicates. A total of 54 articles fulfilled the inclusion criteria and progressed to full-text analysis (Fig. 1). A descriptive summary of reported results was presented in data charting tables (see Appendix C).

PRISMA flow diagram showcasing the selection of evidence for the systematic review ‘DL models available for assisted detection of knee pathologies in MRI’

Among 54 studies analysed, 94% (n = 51) [14, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66] were retrospective cohort studies while comparative analyses made up 4% (n = 2) [67, 68] and diagnostic test accuracy studies made up 2% (n = 1) [69] of the selected studies. Anterior cruciate ligament (ACL) injuries were the focus of 35% (n = 19) of articles [14, 19, 24, 32, 34, 35, 37, 42, 45, 47, 48, 52, 53, 57, 58, 62, 64, 65], followed by meniscal pathologies in 24% (n = 13) [20, 22, 27, 28, 33, 40, 41, 46, 55, 56, 59, 61, 70], general abnormalities in 20% (n = 11) [23, 26, 30, 31, 39, 43, 54, 60, 66,67,68], and osteoarthritis (OA) or cartilage pathologies in 17% (n = 9) [17, 18, 21, 25, 36, 38, 48, 50, 51]. Synovial fluid detection (n = 1, 2%) [29], and tibial fractures (n = 1, 2%) [44] were less frequent. More than half of the studies (n = 29, 54%) used local or private databases [19, 22,23,24, 27, 29, 32,33,34,35,36,40,41,42, 44, 48,49,50,51,52, 55, 58, 59,