This work presents a 70-nW QRS detector implemented in 0.45-V 0.1S-\mu m CMOS technology. The algorithm is optimized with decimated quadratic spline wavelet transform (DQSWT) and window-based extrema difference curve (WEDC). The DQSWT achieves up to 60% computational unit reduction when compared to the conventional method, while the WEDC enhances the QRS detection accuracy while maintaining a low computational complexity. Combined with a decision-making stage with simple mechanism, the proposed algorithm achieves a sensitivity of 99.62% and a precision of 99.70% in MIT-BIH arrhythmia database.