Myocardial pathology segmentation in cardiac magnetic resonance (CMR) images is crucial for diagnosing and managing cardiovascular diseases. However, the challenge of partial modality availability in multi-sequence datasets complicates this task. In this paper, we propose a simple but efficient approach that combines a two-stage prediction strategy with a multi-model ensemble and Region Specific Normalization (RSN) to improve segmentation accuracy. Our method leverages the strengths of models trained on different combinations of CMR sequences, resulting in significant improvements in the Dice scores for myocardial scars and edema. RSN enhances contrast within regions of interest, reducing the impact of artifacts and intensity variability, thereby refining segmentation outcomes. We evaluate our approach using the CARE 2024 MyoPS++ challenge dataset, which includes data from multiple centers with diverse modality availability. The results demonstrate that our method outperforms full-sequence models, both quantitatively and qualitatively, providing a robust and effective solution for myocardial pathology segmentation in clinical settings. Our method achieved second place in the CARE 2024 MyoPS++ challenge, highlighting its efficacy and competitiveness. Code and model are available: https://github.com/Zihaoluoh/CARE-2024_MyoPS- .