Pathology Image Retrieval by Block LBP Based pLSA Model with Low-Rank and Sparse Matrix Decomposition - Advances in Image and Graphics Technologies - page 332

will be converted to the topic-level representation learned from correlated and specific
word-level distributions. After computing the similarities between ROI image and the
images stored in the database, the top R similar images along with the confirmed di-
agnosis information are returned to the CAD system. By comparing ROI with these
returned images, pathologists can make a more reliable diagnosis decision.

Fig. 3. The flow chart of our retrieval framework

4

Experiment

Our proposed method is evaluated on the pathology image database for breast cancer
with confirmed diagnosis information, which is from Motic digital slide database for
the yellow race [20]. The image database consists of 5 categories and 600 images
(256×256, 20x magnification) for each category, as shown in Fig. 4.

Fig. 4. Five categories of digital pathology slides. (a) Basal-like carcinoma (BLC). (b) Breast
myofibroblastoma (BMFB). (c) Invasive breast cancer (IBC). (d) Low-grade adenosquamous
carcinoma (LGASC). (e) Mucinous cystadenocarcinoma (MCA).