A digital mouse atlas, extracted from CT and cryosection data,33 was utilized to perform the gastric cancer-mimic simulation to validate the ability of the HRDM-based algorithm in an irregular shape phantom simulation. In the simulation, only the torso section of the digital mouse was selected, and the detailed optical properties of each organ are listed in Table 2.34 In this study, the stomach was modeled as an air bubble, termed as the gastric pouch, which was used to simulate the cavity organ. Because it is difficult to localize the stomach wall during organ segmentation and impossible to discretize it during finite element discretization, the stomach wall was not modeled. An ellipsoid with the size of (1, 1.5, 1) mm was positioned at (26, 12, 23) mm to mimic the inner bioluminescent probe that has been marked on the gastric cancer cells, as shown in Fig. 2(a). Similar to the cervical cancer-mimic simulation, the power of the source was set to be the 1 nW for simplicity, and the power density was calculated as . The measured light flux distribution on the surface of the digital mouse was simulated by MOSE, as shown in Fig. 2(b). The reconstructed results for both the HRDM-based and the DE-based algorithms were obtained and are presented in Fig. 3, where Fig. 3(a)–3(c) presents the coronal, axial and sagittal view results of the HRDM based algorithm, and Fig. 3(d)–3(f) shows those for the DE-based algorithm. The pink sphere represents the actual source, and the colored domains around it are the reconstructed sources. The detailed quantitative comparison results are listed in Table 3. Similarly, more accurate and acceptable results were obtained by the HRDM-based algorithm than the DE method. The distance error of the HRDM was 0.63 mm, which was smaller than the size of the actual source and much better than that of the DE (3.05 mm). On the other hand, a more accurate quantification result was obtained by the HRDM, in which the reconstructed source power density was with the quantification error of 5.88%. The reconstructed source power density of the DE was overestimated (greater than 150%) because of neglecting the non-scattering region. The comparison results demonstrated the accuracy of the HRDM-based algorithm in the irregular shape phantom, and also revealed its superiority to the DE-based algorithm when the non-scattering region cannot be neglected.