Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Axial view MRI of L5-S1 IVD showing a) central stenosis and b) left and right lateral foraminal stenosis, both of which impinging upon the nerve bundle contained in the thecal sac and the nerve roots exiting the spine.

More »

Fig 1 Expand

Fig 2.

The three measurements that are necessary to determine the extent of LSS.

The a) AP diameter is used to determine how much pressure exerted on the spinal cord, b) left and c) right foraminal widths which are used to determine the pressure exerted on left and right spinal nerve roots, respectively. The red and blue lines delineate the important boundaries of IVD and PE regions, respectively.

More »

Fig 2 Expand

Fig 3.

An overview of the overall methodology.

More »

Fig 3 Expand

Fig 4.

An example of the T1- and T2-weighted composite input image (left) and the label image used to train the SegNet model (right). The regions in the label image are 1) Unregistered, 2) IVD, 3) PE, 4) TS, 5) AAP and 6) Other.

More »

Fig 4 Expand

Fig 5.

A color-coded boundary of a manually labeled lumbar spine MRI image (left) and the automatic segmentation result (right). The left image clearly shows ambiguities and inaccuracies, particularly along the important boundary locations.

More »

Fig 5 Expand

Fig 6.

An example of a 4x4 image (top left) that is segmented into a label image containing three regions with red lines marking the boundaries between them (top right) and its Boundary Grid (bottom).

More »

Fig 6 Expand

Fig 7.

The horizontal gradient image (top left) and the vertical gradient image (top right) and the Modified Boundary Grid (bottom) of the 4x4 input and label images shown at the top of Fig 6.

More »

Fig 7 Expand

Table 1.

Sparse boundary representation.

More »

Table 1 Expand

Fig 8.

An overview of the proposed contour evolution process.

More »

Fig 8 Expand

Fig 9.

The location of nine important points for determining AP diameter and foraminal widths.

More »

Fig 9 Expand

Table 2.

Averaged BF-score of important IVD and PE boundaries after improvement.

More »

Table 2 Expand

Fig 10.

A typical example of the number of pixel changes between successive iterations of our proposed method.

More »

Fig 10 Expand

Fig 11.

A typical example of the number of pixel changes between successive iterations of the geodesic active contours method.

More »

Fig 11 Expand

Fig 12.

A typical example of the number of pixel changes between successive iterations of the Chan-Vese method.

More »

Fig 12 Expand

Fig 13.

An example of the result of the application of a contour evolution method to the a) the original boundary using b) our proposed method, c) Geodesic Active Contours and d) Chan-Vese methods. The numbers mark the location of the four cases that illustrate the superiority of our method to the others.

More »

Fig 13 Expand

Table 3.

The intra-expert variations, ∨, calculated as the average distance, in mm, between each of the five sample data points and their mean locations.

More »

Table 3 Expand

Table 4.

The parameters of each experiment setup.

More »

Table 4 Expand

Table 5.

Mean error in mm, ε, between each predicted important point location and its reference location (column 2–7), between the predicted left and right foraminal widths and AP diameter and their corresponding values (column 8–10) and the IVD herniation diagnosis accuracy in % (last column).

More »

Table 5 Expand

Fig 14.

The visual interface of the developed computer-aided diagnostic system showing the results boundary delineation results, foraminal widths measurements, and central IVD herniation classification.

The figure shows a) a healthy L3-L4 disk, b) and c) IVDs with different types of abnormalities and d) a case where the right foraminal width measurement and diagnostic results are erroneous.

More »

Fig 14 Expand