Method and device of detecting bone mineral density based on low dose computed tomography image of chest

1 . A method of processing a low-dose computed tomography (LDCT) chest image, comprising:

receiving a first set of chest images, the first set of chest images generated by a low-dose CT method;

determining a vertebral body region in each image of the first set of chest images;

determining an anterior part within each vertebral body region;

obtaining a first set of features from the anterior parts of the first set of chest images;

determining a curved graph based on a sequence of the first set of chest images versus the first set of features of the anterior parts of the vertebral body regions of the first set of chest images;

selecting a second set of features from the first set of features; and

determining whether a bone mineral density of the vertebral body regions is abnormal based on the second set of features.

2 . The method of claim 1 , wherein whether the bone mineral density of the vertebral body regions is abnormal is determined by a first classification model.

3 . The method of claim 2 , further comprising:

when the bone mineral density of the vertebral body regions is abnormal, determination of whether the bone mineral density of the vertebral body regions belongs to osteopenia or osteoporosis by a second classification model.

4 . The method of claim 1 , wherein the vertebral body region in each of the first set of chest images is determined by a first segmentation model.

5 . The method of claim 1 , wherein the anterior part of the vertebral body region is determined by a second segmentation model.

6 . The method of claim 1 , wherein the vertebral body region includes thoracic spine, vertebral body, intervertebral disc, cortical bone, basivertebral vein, endplate, and trabecular bone.

7 . The method of claim 6 , wherein the anterior part of the vertebral body region includes the trabecular bone.

8 . The method of claim 7 , wherein the anterior part of the vertebral body region does not include the cortical bone and the basivertebral vein.

9 . The method of claim 1 , wherein the second set of features is selected by an anomaly detection model or a third classification model.

10 . The method of claim 1 , wherein the second set of features is selected based on first differential values or second differential values of the curved graph.

11 . The method of claim 1 , wherein the second set of features correspond to the anterior parts including a trabecular bone.

12 . The method of claim 1 , further comprising:

selecting a second subset of chest images from the first set of chest images, the second subset of chest images corresponding to the second set of features; and

determining whether the bone mineral density of the vertebral body regions is abnormal based on the second subset of chest images.

13 . The method of claim 12 , further comprising:

determining a vector based on the second subset of chest images; and

determining whether a bone mineral density of the vertebral body regions is abnormal based on the vector.

14 . The method of claim 1 , wherein, when the bone mineral density of the vertebral body regions is abnormal, the method further comprises:

determining a bone mineral density level based on the second set of features; and

outputting the bone mineral density level.

15 . The method of claim 1 , wherein the anterior part of the vertebral body region is determined by shrinking an area of the vertebral body region in each of the first set of chest images.

16 . The method of claim 15 , wherein shrinking the area of the vertebral body region in each of the first set of chest images comprises shrinking 2 to 3 pixels from the boundary of the area of the vertebral body region.

17 . The method of claim 1 , wherein the second set of the features selected from the first set of features exclude a plurality of images, the plurality of images including an endplate and an intervertebral disc.

18 . The method of claim 1 , further comprising:

determining a lung region in each image of the first set of chest images;

for each image of the first set of chest images, determining a ratio of an area of the lung region to the whole image;

discarding images from the first set of chest images, in which each of the discarded images have the ratio smaller than a first threshold.

19 . A device of processing a low-dose computed tomography (LDCT) chest image, comprising:

a processor; and

a memory coupled with the processor,

wherein the processor executes computer-readable instructions stored in the memory to perform operations, and the operations comprise:

receiving a first set of chest images, the first set of chest images generated by a low-dose CT method;

determining a vertebral body region in each image of the first set of chest images;

determining an anterior part within each vertebral body region;

obtaining a first set of features from the anterior parts of the first set of chest images;

determining a curved graph based on a sequence of the first set of chest images versus the first set of features of the anterior parts of the vertebral body regions of the first set of chest images;

selecting a second set of features from the first set of features; and

determining whether a bone mineral density of the vertebral body regions is abnormal based on the second set of features.