<p>An electrocardiogram (ECG) signal is a fundamental tool in cardiology to identify different heart-related diseases. Various noises contaminate the ECG signal during acquisition, thereby attenuating the accuracy of diagnoses. This article proposes a denoising technique to minimize interference in the ECG signal using an extended Kalman smoother (EKS) framework with a modified lightning search algorithm (MLSA). The MLSA optimizes ten elements of the input signal to create a synthetic signal similar to a real-time ECG signal. The EKS framework utilizes the optimized parameters to form a state and measurement equation to eliminate interference in the ECG signal. The developed approach (EKS + MLSA) is analyzed by adding different types of noise to the MIT-BIH arrhythmia database at various signal-to-noise ratios (SNRs). The output of the EKS + MLSA is identical to a pure ECG signal for denoising various noise-added ECG signals. The performance of EKS + MLSA is evaluated in results in terms of SNR improvement, mean square error (MSE), mean absolute error (MAE), correlation coefficient (CC), and percentage of root mean square distortion (PRD) for the removal of various noises from the noise-added ECG signal at different SNRs. The developed EKS + MLSA technique outperforms other noise cancellation methods, such as wavelet denoising (WD), non-local mean (NLM), empirical mode decomposition (EMD) with NLM, and variational mode decomposition (VMD) techniques in terms of visual quality and evaluation metrics.</p>