Purpose Autologous subchondral bone grafting in combination with autologous chondrocyte implantation (ACI) (sandwich procedure) is a well-accepted procedure for the treatment of osteochondral lesions of the knee. This requires a different surgical technique and preoperative planning compared to ACI alone. In addition, pain from bone marrow donor site locations can be expected and should be part of patient consent and expectations. This study evaluates whether the MRI made as part of the standard preoperative cartilage patient work up has the diagnostic accuracy to predict the need for a sandwich procedure. Methods and materials Retrospectively, 185 preoperative MRI scans (PD and T2 sequences) of patients planned for ACI were included. The integrity of the subchondral bone and lamina was scored by four different observers (3 radiologists, and 1 orthopaedic resident). The depth of the defect was measured perpendicular from articulating surface to the bottom of the bony lesion. The area under the curve (AUC) for subchondral defect on MRI (i.e. lamina or bone defect or expert impression), depth measurements and eventual sandwich procedure were calculated. Also inter-observer Kappa values were determined. Results The AUCs for lamina (0.74–0.80) and bone defect (0.73–0.79) were fair and inter-observer Kappas ranged from 0.49 to 0.76, indicating a moderate-good inter-observer agreement and moderate prediction of the need for a sandwich procedure based on the presence of lamina and or subchondral bone defect on MRI. However, depth measurements resulted in an AUC of 0.90 (95% CI: 0.84–0.95,) with an optimal cut-off point at 6.5 mm depth of the lesion (90% sensitivity, 80% specificity) to predict the need for a sandwich procedure. Conclusion Ours is the first study examining MRI as a diagnostic tool in predicting the need for a sandwich procedure. Our results show that the integrity of the subchondral layer on MRI has a moderate role in predicting the need for an eventual autologous bone graft to augment ACI whereas in our cohort a depth of the lesion above 6.5 mm accurately predicts the need for a sandwich procedure. This can aid in optimising the preoperative planning and patient consent.