An optimized approach to nonlinear iterative reconstruction of magnetic resonance imaging (MRI)–guided near-infrared spectral tomography (NIRST) images was developed using an L-curve-based algorithm for the choice of regularization parameter. This approach was applied to clinical exam data to maximize the reconstructed values differentiating malignant and benign lesions. MRI/NIRST data from 25 patients with abnormal breast readings (BI-RADS category 4-5) were analyzed using this optimal regularization methodology, and the results showed enhanced values and area under the curve (AUC) for the task of differentiating malignant from benign lesions. Of the four absorption parameters and two scatter parameters, the most significant differences for benign versus malignant were total hemoglobin (HbT) and tissue optical index (TOI) with and 0.001, and AUC and 0.94, respectively, in terms of HbT and TOI. This dramatically improved the values relative to fixed regularization ( and 0.003; and 0.83) showing that more differentiation was possible with the optimal method. Through a combination of both biomarkers, HbT and TOI, the AUC increased from 82.9% () to 94.3% (optimal method).