Accurate segmentation of retinal cysts in Optical Coherence Tomography (OCT) images is critical for diagnosing and monitoring diseases like Diabetic Macular Edema. However, progress in automated methods is hindered by a lack of specialized public datasets. To address this, we introduce a new, meticulously annotated OCT dataset for cyst segmentation. To the best of our knowledge, it is the first dataset to be presented with a predefined, patient-based split for training, validation, and testing, ensuring a realistic evaluation of model generalization. We establish performance benchmarks using U-Net and its variants. Our results provide a solid baseline, revealing that while larger cysts are well-detected, segmenting small-scale cysts among speckle noise remains a key challenge for current models. This dataset and its corresponding benchmarks will serve as an important resource to accelerate the development and standardized evaluation of more robust and clinically relevant segmentation algorithms.