Desktop version

Home arrow Health

Examples of Lymph Node Detection on Abdominal CT Images

This section will show examples of the results of lymph node detection from abdominal CT images. Acquisition parameters of these CT images are image size

Flowchart of false-positive reduction based on support vector machine

Fig. 3.13 Flowchart of false-positive reduction based on support vector machine

Example of lymph node regions detected by Hessian-based analysis. This figure shows examples of detected true-positive regions

Fig. 3.14 Example of lymph node regions detected by Hessian-based analysis. This figure shows examples of detected true-positive regions

512 x 512 x 401-451 voxels, pixel size 0.586-0.702 mm2, section thickness 1.25 mm, and reconstruction pitch 0.5-1.0 mm. The data from 28 cases of contrast- enhanced three-dimensional (3D) abdominal CT examinations were processed. These image datasets included five colorectal cancers and 23 gastric cancers. There were 95 lymph nodes included in these images. The parameters for detection were set as Tbse = 8 and Tx3 = 1.1.

Figure 3.14 shows examples of true-positive regions extracted by the method shown in Sect. 3.4.2. As shown in this figure, Hessian-based analysis can detect lymph nodes in the pelvis appropriately. Although the method can detect truepositive regions from 3D abdominal CT, it also detected several false-positive regions and false-negative regions (Figs. 3.15 and 3.16). As we can see on these

Example of regions detected by Hessian-based lymph node detection method. This figure shows examples of false-positive regions

Fig. 3.15 Example of regions detected by Hessian-based lymph node detection method. This figure shows examples of false-positive regions

Example of lymph nodes missed by automated method

Fig. 3.16 Example of lymph nodes missed by automated method

figures, other anatomical regions such as a part of the colonic wall or residues inside the intestine tended to be extracted as false-positive regions. The overall performance of detection is shown in Fig. 3.17 as a receiver operating characteristics (ROC) curve.

Receiver operating characteristic (ROC) curve of lymph node detection method

Fig. 3.17 Receiver operating characteristic (ROC) curve of lymph node detection method

 
Source
< Prev   CONTENTS   Source   Next >

Related topics