Scintillators and photoconductors used in energy integrating detectors (EIDs) have inherent variations in their imaging response to single-detected X-rays due to variations in X-ray energy deposition and secondary quanta generation and transport, which degrades DQE(f). The imaging response of X-ray scintillators to single X-rays may be recorded and studied using single X-ray imaging (SXI) experiments; however, no method currently exists for relating SXI experimental results to EID DQE(f). This work proposes a general analytical framework for computing and analyzing the DQE(f) performance of EIDs from single X-ray image ensembles using a spatial frequency-dependent pulseheight spectrum. Methods: A spatial frequency (f)-dependent gain, g(f ), is defined as the Fourier transform of the imaging response of an EID to a single-detected X-ray. A f-dependent pulse-height spectrum, Pr[ g(f )], is defined as the 2D probability density function of g(f ) over the complex plane. Pr[ g(f )] is used to define a f-dependent Swank factor, A S (f), which fully characterizes the DQE(f) degradation due to single X-ray noise. A S (f) is analyzed in terms of its degradation due to Swank noise, variations in the frequency-dependent attenuation of | g(f )|, and noise in arg g(f ) which occurs due to variations in the asymmetry in each single X-ray's imaging response. Three example imaging systems are simulated to demonstrate the impact of depth-dependent variation in g(f ), remote energy deposition, and a finite number of secondary quanta, on Pr[ g(f )], A S (f), MTF(f),and NPS(f)/NPS(0),which are computed from ensembles of single X-ray images. The same is also demonstrated by simulating a realistic imaging system; that is, a Gd 2 O 2 S-based EID. Using the latter imaging system, the convergence of A S (f) estimates is investigated as a function of the number of detected X-rays per ensemble. Results: Depth-dependent g(f ) variation resulted in A S (f) degradation exclusively due to depth-dependent optical Swank noise and the Lubberts effect. Conversely, the majority of A S (f) degradation caused by remote energy deposition and finite secondary quanta occurred due to variations in arg g(f ). When using input X-ray energies below the K-edge of Gd, variations in the frequencydependent attenuation of | g(f )| accounted for the majority of A S (f) degradation in the GOS-based EID, and very little Swank noise and variations in arg g(f ) were observed. Above the K-edge, however, A S (f) degradation due to Swank noise and variations in arg g(f ) greatly increased. The convergence of A S (f) was limited by variation in arg g(f ); imaging systems with more variation in arg g(f ) required more detected X-rays per ensemble.