摘要：Three-dimensional (3D) image reconstruction of tumors based on serial histological sectioning is one of the most powerful methods for accurate high-resolution visualization of tumor structures. However, 3D histological reconstruction of whole tumor has not yet been achieved. We established a high-resolution 3D model of molecular marked whole laryngeal cancer by optimizing the currently available techniques. A series of 5,388 HE stained or immunohistochemically stained whole light microscopic images (200 ×) were acquired (15.61 TB).The data set of block-face images (96.2 GB) was also captured. Direct volume rendering of serial 6.25 × light microscopy images did not demonstrate the major characteristics of the laryngeal cancer as expected. Based on fusion of two datasets, the accurate boundary of laryngeal tumor bulk was visualized in an anatomically realistic context. In the regions of interest, micro tumor structure, budding, cell proliferation and tumor lymph vessels were well represented in 3D after segmentation, which highlighted the advantages of 3D reconstruction of light microscopy images. In conclusion, generating 3D digital histopathological images of a whole solid tumor based on current technology is feasible. However, data mining strategy should be developed for complete utilization of the large amount of data generated.
其他摘要：Abstract Three-dimensional (3D) image reconstruction of tumors based on serial histological sectioning is one of the most powerful methods for accurate high-resolution visualization of tumor structures. However, 3D histological reconstruction of whole tumor has not yet been achieved. We established a high-resolution 3D model of molecular marked whole laryngeal cancer by optimizing the currently available techniques. A series of 5,388 HE stained or immunohistochemically stained whole light microscopic images (200 ×) were acquired (15.61 TB).The data set of block-face images (96.2 GB) was also captured. Direct volume rendering of serial 6.25 × light microscopy images did not demonstrate the major characteristics of the laryngeal cancer as expected. Based on fusion of two datasets, the accurate boundary of laryngeal tumor bulk was visualized in an anatomically realistic context. In the regions of interest, micro tumor structure, budding, cell proliferation and tumor lymph vessels were well represented in 3D after segmentation, which highlighted the advantages of 3D reconstruction of light microscopy images. In conclusion, generating 3D digital histopathological images of a whole solid tumor based on current technology is feasible. However, data mining strategy should be developed for complete utilization of the large amount of data generated.
