Obstructive Sleep Apnea (OSA) is a common sleep disorder that can significantly impact health if not diagnosed or treated. Traditionally, detecting sleep apnea using Polysomnography (PSG) includes complex, costly, and time-consuming activities. Whereas deep learning-based automatic detection can overcome these issues. This study presents a one dimensional Convolutional Neural Network Long Short Term Memory (1D CNN-LSTM) architecture for automatic detection of sleep apnea using ECG signal. The ECG signals provide vital information about heartbeats, which are closely associated with respiratory patterns during sleep. The signals were preprocessed by filtering and normalization techniques to improve the quality. A comprehensive collection of 41 features representing- time, frequency and non-linear domain, were extracted to capture detailed variation linked to apnea events. The CNN layers extract spatial characteristics while the LSTM layers learn temporal dependencies, which is crucial for classification. The model was evaluated on the PhysioNet Apnea-ECG dataset, using 10-fold cross-validation and achieved a high accuracy of 89.14%, specificity of 86.49%, sensitivity of 91.79%, F1-score of 89.42% and AUC-ROC of 0.9553, with minimal overfitting. With its strong performance and low computational cost, the suggested method can easily be implemented in real-world OSA detection applications and can serve as a foundation for future enhancements.