Figure 9
From: Novel coarse-to-fine dual scale technique for tuberculosis cavity detection in chest radiographs
An example of cavity candidates detection using the proposed technique. (a) Original CXR image; (b) candidate detection results in ULZ obtained using GTM + LHFC where the green rectangular windows (C1, C2, C3) represent the candidates, and the blue dotted contour is the true cavity annotated by radiologists; (c) magnified candidate windows: C1–C3 (left to right); (d) HIE results of C1–C3; (e) improved fluid vector flow (IFVF) results of C1–C3 with the help of HIE; (f) final cavity detection results using fine feature classification. Red contour is the detected cavity, while the cyan ones are the non-cavity contours.