Transverse relaxometry, a quantitative MRI technique that measures the signal decay time (T₂), has shown promise in detection of subtle abnormalities in neurological diseases¹. However, historically, this technique has produced inconsistent results, due primarily to sub-optimal data fitting. Quantitative T₂ mapping is highly sensitive to transmit field heterogeneities (B₁⁺) ², and in the case of slice selective imaging, to radio-frequency (RF) pulse shapes which cause imperfections in signals generated with a multi-echo spin echo sequence. Conventional fitting methods ignore transmit imperfections (and the resulting echo oscillations). One strategy is to simply discard early echo times to improve fitting³, but this risks compromising tissue characterization⁴. A recently proposed fitting method, called stimulated echo correction (SEC)⁵, estimates major confounds associated with fitting errors in the transmit field and returns less biased results. SEC is a one-step least square method in which 3 parameters (T₂, amplitude, and B₁⁺) are estimated. While T² and amplitude are not necessarily guaranteed to have strong spatial correlations, the B₁⁺ field is not expected to vary rapidly and can be highly constrained. Our aim is to develop and employ an improved SEC fit (called iSEC) that constrains B₁⁺ in order to reduce inconsistencies associated with T₂ estimation.Our proposed iSEC method is a two-step procedure: the first pass consists of the standard SEC method to provide an initial estimate of T₂, amplitude, and B₁⁺. The B₁⁺ field is then spatially filtered by convolving with a Gaussian window with 3mm FWHM. The second pass uses the filtered B₁⁺ maps as a pre-estimated input and re-fits the other 2 parameters (T₂ and amplitude). We compared and investigated reliability between the proposed iSEC and the standard SEC fit with simulated and in-vivo data. Simulated data were generated with the extended phase graph algorithm assuming nominal acquisition and tissue parameters: 16 echoes, 10 ms echo spacing, T₁/T₂ of 3000/100 ms, and B₁⁺ reduced to 0.75 of its ideal value. Gaussian noise (10<SNR<100) was added to mimic real data; 1000 noise realizations were performed at each SNR value. T2 values were estimated using the original SEC and our proposed iSEC algorithm. Experimental verification was performed by acquiring 10 sequentially repeated scans in one volunteer. The sequence consisted of a multi-echo spin echo with 12 slices, 16 echoes, and 9.2 ms echo spacing on a 3T GE MR750 scanner T₂ maps generated with both methods were evaluated based on variance across trials.Simulations indicate that both SEC and iSEC methods provide equally accurate T₂ estimation over a wide range of SNR, Figure 1. However, the restricted B₁⁺ field in the iSEC method translates to a reduced variance at all SNR values: T₂ values estimated with iSEC had ~25% lower standard deviation than those estimated with SEC. iSEC was found to be particularly beneficial in low SNR regions (<35) as well as within regions with low B₁⁺, Figure 2. With the nominal parameters used in the simulation, tissues with T₂ values below approximately 110 ms (corresponding to white and gray matter ⁶) benefited the most with our proposed fit, Figure 3. Our experimental MR images also produce similar results. Estimated T₂ maps illustrate relatively similar average T₂ values for regions with high SNR (> 50) for both SEC and iSEC. Repeated scans indicate that standard deviations within the regions with low B1+ fraction (about 0.75) are lowered 24% in iSEC compared to SEC, Figure 4. In regions with near-ideal B1+ (above 0.9), the T₂ sensitivity to mis-estimates of B₁⁺ are minor and less than 10% reduction in variance little benefit is observed. Overall, simulated and experimental results suggest that a B₁⁺ constrained fit is able to improve fits under all circumstances but is especially beneficial when the transmit field is low.Simulated and experimental data indicate that precision in T₂ estimation is always improved with iSEC relative to SEC. The improvement observed with iSEC is greatest in low SNR regions and in situations when system imperfections are severe. This translates to reliable T₂ estimations with thinner slices or with higher spatial resolution. This may also be important if the refocusing angles are intentionally lowered to reduce SAR or to achieve shorter echo spacing. This situation will mimic an unintentionally low B₁⁺ field, where we observe benefit in a B₁⁺ constrained fit. Ultimately, iSEC is expected to translate into more reliable T₂ maps than can currently be generated, which may provide reliable detection of pathology in neurological disorders particularly temporal lobe epilepsy in which extensive advantages of T₂ relaxometry-aided diagnoses have already been shown¹.