Diffuse　optical　tomography　(DOT)　is　a　non-invasive,　label-free　imaging　technique　widely　used　in　applications　such　as　breast　cancer　diagnosis　and　brain　imaging.　It　allows　for　the　quantitative　measurement　of　tissue　functional　parameters,　including　the　concentrations　of　oxyhemoglobin,　deoxyhemoglobin,　and　water.　However,　the　quality　of　reconstructed　images　is　poor　due　to　light　scattering.　To　address　this　challenge,　one　effective　strategy　is　to　incorporate　anatomical　information　from　high-resolution　imaging　to　guide　DOT　reconstruction.　In　this　study,　a　new　approach　(GRI)　is　developed　to　leverage　MRI　images　based　on　graph　structure　for　DOT　reconstruction.　And　its　feasibility　and　effectiveness　were　evaluated　with　numerical　simulations.　The　results　indicate　that　GRI　significantly　improves　the　structural　similarity　index　(SSIM)　of　reconstructed　total　hemoglobin　(HbT)　images,　outperforming　conventional　Tikhonov　regularization　and　direct　regularization　imaging　(DRI)　method　by　more　than　63%　and　82%,　respectively.　These　findings　are　further　supported　by　experiments　using　real　patient　data,　which　demonstrate　GRI＇s　potential　in　breast　cancer　diagnosis.　©　2025　SPIE.