Purpose The aims were to use magnetic resonance imaging (MRI) to compare the efficacy of fat-suppressed proton-density turbo-spin-echo (FS-PD-TSE) images and T1-weighted (T1WIs) and T2-weighted images (T2WIs) in identifying cartilaginous endplate failure (CEF), and to propose a modified Rajasekaran classification based on the FS-PD-TSE sequence. Methods Thirty-one lumbar disc herniation (LDH) cases were enrolled. Totally, 155 discs and 310 endplates were evaluated by MRI with T1W, T2W, and FS-PD-TSE sequences. Disc degeneration (DD), LDH grades, and the total endplate score (TEPS) of CEF were evaluated. Chi-square, Spearman rank correlation analysis, and multiclass logistic regression were used to compared the sensitivity in diagnosing CEF. A modified Rajasekaran classification based on FS-PD-TSE sequence was established to diagnose CEF. The multiclass logistic regression model was used to analyse the relationship between modified Rajasekaran classification and DD and LDH. Results There were 117 (75.5%) segments with CEF in T1WIs, 115 (74.2%) in T2WIs, and 127 (81.9%) in FS-PD-TSE, respectively. Chi-square test showed FS-PD-TSE images were more sensitive than T1WIs and T2WIs (P < 0.05). Spearman rank correlation analysis revealed a significant correlation between TEPS and LDH and DD in T1WIs, T2WIs, and FS-PD-TSE images (P < 0.05). A multiclass logistic regression model showed that the incidence of DD and LDH significantly increased accordingly with increases in modified Rajasekaran classification (P < 0.05). ConclusionThe FS-PD-TSE sequence has high diagnostic value for lumbar CEF. CEF is a risk factor for LDH. The new classification for lumbar CEF based on the FS-PD-TSE sequence has good predictive ability for LDH and DD.