Background: Whole-exome sequencing (WES) is a strong diagnostic tool for foetal structural abnormalities, but the causative gene for more than half abnormalities has not been identified. Therefore, it is essential to improve the diagnostic yield based on WES data. Methods: First, 138 foetuses with structural abnormalities were subjected to conventional WES and copy number variation (CNV) analyses. For undiagnosed cases, we employed a three-step approach for diagnosis. 1) We re-evaluated candidate variants using a loss-of-function observed/expected upper bound fraction (LOEUF) score, 2) we re-evaluated all variants on disease-causing genes for clinically diagnosed cases using spliceAI, and 3) we re-evaluated the rare variants in all low LOEUF scored genes (< 0.35) using spliceAI. Results: We identified molecular diagnoses in 53 of the 138 cases (38.4%) using conventional WES and CNV analysis. For undiagnosed cases, using the first step, we diagnosed two X-linked recessive diseases. Using the second step, we diagnosed Meckel-Gruber syndrome by detecting likely pathogenic intron variant in TMEM67. In the third step, we identified a de novo hemizygous pathogenic variant in one severe hydrops fetalis male, which could cause aberrant splicing in CASK. We found a novel phenotype, hydrops fetalis, in CASK-related X-linked dominant disorder. Moreover, we revealed that the LOEUF score of X-linked disease-causing genes was significantly lower than that of autosomal genes among all OMIM-registered genes. Conclusion: We showed that the evaluation of variants, including introns of WES data, in combination with the LOEUF score, could improve the WES diagnostic yield and detect new candidate genes, especially on chromosome X.