阅读说明：本技术 使用双能量计算机断层扫描的针对外伤性出血的检测和量化 (Detection and quantification of traumatic bleeding using dual energy computed tomography ) 是由 徐宙冰 S.格尔比奇 周少华 P.赫尔策尔 G.索扎 于 2019-09-06 设计创作，主要内容包括：公开了使用双能量计算机断层扫描的针对外伤性出血的检测和量化。提供用于针对外伤性出血进行自动检测和量化的系统和方法。使用全身双能量CT扫描仪来获取图像数据。经机器学习的网络检测在来自双能量CT扫描图像数据的出血图上的一个或多个出血区域。从出血图生成显像。对所预测的出血区域进行量化,并且生成风险值。将显像和风险值呈现给操作者。(Detection and quantification of traumatic bleeding using dual energy computed tomography is disclosed. Systems and methods for automated detection and quantification of traumatic bleeding are provided. Image data is acquired using a whole-body dual-energy CT scanner. The machine-learned network detects one or more bleeding regions on a bleeding map from dual-energy CT scan image data. An image is generated from the bleeding map. The predicted bleeding area is quantified and a risk value is generated. The visualization and risk values are presented to the operator.)

1. A method for predicting an area of bleeding in a patient, the method comprising:

acquiring dual-energy image data using a dual-energy CT whole-body scan of a patient;

predicting one or more bleeding areas on a bleeding area map using a neural network trained to identify and classify bleeding areas given dual-energy image data;

generating a visualization of the one or more bleeding areas; and

displaying the display.

2. The method of claim 1, wherein the dual energy image data comprises at least high kV CT image data and low kVCT image data.

3. The method of claim 1, further comprising:

synthesizing virtual monochrome image data from the dual-energy image data;

wherein the synthesized virtual monochrome image data is used to detect the bleeding area.

4. The method of claim 1, further comprising:

wherein the dual energy image data comprises single energy image data having different keV levels.

5. The method of claim 1, further comprising:

the dual energy image data is processed to remove bone material and to track blood vessels.

6. The method of claim 5, wherein processing comprises:

identifying a material type in the dual-energy image data using the material map;

segmenting the dual-energy image data using the material type;

masking bone material using the segmented dual energy image data; and

the segmented dual energy image data is used to track blood vessels.

7. The method of claim 1, wherein the neural network is a convolutional neural network.

8. The method of claim 7, wherein the neural network is trained using manually annotated whole-body images of bleeding patients.

9. The method of claim 1, further comprising:

identifying one or more regions of interest in the dual-energy image data;

inputting image data for the one or more regions of interest into a neural network in place of dual energy image data.

10. The method of claim 1, further comprising:

quantifying the severity of the one or more bleeding areas; wherein the visualization is generated with a quantified indication for the one or more bleeding areas.

11. A method for automatically assessing traumatic bleeding, the method comprising:

acquiring DECT image data of a patient using a dual energy CT scan;

removing bone material from the DECT image data;

tracking blood vessels in DECT image data;

predicting one or more bleeding regions using a first machine-learned network trained to identify and classify bleeding regions given DECT image data;

determining a severity rating for the one or more bleeding areas;

calculating a size for the one or more bleeding areas;

generating a rating for the one or more bleeding areas as a function of a size and severity rating of the one or more bleeding areas;

estimating a risk value for the patient using a second machine-learned network trained to express an overall risk; and

outputting the risk value.

12. The method of claim 11, wherein determining a severity rating comprises:

generating a severity map using landmark detection and historical clinical reports; and

a severity map is used to rank the severity of each predicted bleeding area.

13. The method of claim 11, wherein removing bone material comprises:

identifying bone material in the image data using the material map;

segmenting the image data; and

bone material in the segmented image data is masked.

14. The method of claim 11, wherein calculating the size comprises:

identifying a volume and a diameter for each of the one or more bleeding regions.

15. The method of claim 11, wherein the first machine-learned network includes a deep learning-based classifier.

16. The method of claim 11, further comprising:

outputting a visualization and a risk value for the one or more bleeding areas.

17. A system for automatically assessing traumatic bleeding, the system comprising:

a dual energy computed tomography scanner configured to acquire high kV CT image data and low kV CT image data of a patient;

an image processor configured to input high kV CT image data and low kV CT image data into a machine-learned network trained to predict bleeding regions given image data; the image processor is further configured to generate a visualization of the bleeding region predicted by the machine-learned network; and

a display configured to display the visualization.

18. The system of claim 17, wherein the dual-energy computed tomography scanner is configured to perform a whole-body scan of the patient.

19. The system of claim 17, further comprising:

an assessment module configured to estimate a risk value for the patient using a second machine-learned network trained to express an overall risk given one or more predicted bleeding areas.

20. The system of claim 19, wherein the evaluation module is further configured to estimate a risk value by calculating a size and severity rating for the one or more predicted bleeding areas.

Technical Field

The present embodiments relate to medical computed tomography imaging and diagnosis.

Background

Traumatic bleeding is one of the leading causes of death from accidents. Traumatic bleeding requires immediate treatment and urgent care. More than 50% of all trauma patients with fatal outcome die within 24 hours of injury. Proper management of trauma patients with major bleeding includes early identification of potential sources of bleeding, followed by immediate action to minimize blood loss, restore tissue perfusion, and achieve stabilization. However, detection of traumatic bleeding is very challenging, especially when there are no apparent defects on the skin or skull surface.

Computed Tomography (CT) has previously been used to scan trauma patients. Using a CT scanner, the total whole-body scan time can be reduced to less than 30 seconds. While CT scanners are capable of capturing images of trauma patients, analysis of the resulting data, particularly to identify bleeding, is difficult and inconvenient. Potential bleeding may be located anywhere on the patient (e.g., brain, GI, thorax, abdominal cavity, etc.), and thus bleeding is differentiated from a wide variety of surrounding structures. Further, there is a great variability in the appearance of bleeding. Furthermore, for CT images, it is possible to delineate bleeding regions with similar intensity values to other structures, such as bones and blood vessels, which may complicate the diagnosis region.

Disclosure of Invention

By way of introduction, the preferred embodiments described below include embodiments for automated detection and quantification of traumatic bleeding. Image data is acquired using a whole-body dual-energy CT scan. The machine-learned network generates a bleeding probability map from the dual-energy CT scan. And generating a display according to the bleeding probability map. The predicted bleeding area is quantified and a risk value is generated. The visualization and risk values are presented to the operator.

In a first aspect, a method for detecting a bleeding area in a patient is provided. Dual energy image data is acquired using a dual energy CT whole body scan of a patient. Bleeding areas are detected on a bleeding area map using a neural network trained to identify and classify one or more bleeding areas given dual-energy image data. A visualization of the hemorrhage map is generated and displayed.

In a second aspect, a method for automated assessment of traumatic bleeding is provided. DECT image data of a patient is acquired using a dual energy CT scan. Bone material is removed from the DECT image data. Blood vessels are tracked in DECT image data. The one or more bleeding regions are predicted using a first machine-learned network trained to identify and classify bleeding regions given DECT image data. A severity rating for one or more bleeding areas is determined. The size for one or more bleeding areas is calculated. A rating is generated for the one or more bleeding areas as a function of a size and severity rating of the one or more bleeding areas. Estimating a risk value for the patient using a second machine-learned network trained to express an overall risk.

In a third aspect, a system for automated assessment of traumatic bleeding is provided. The system includes a dual energy computed tomography scanner, an image processor, and a display. A dual energy computed tomography scanner is configured to acquire high KV CT image data and low KV CT image data of a patient. The image processor is configured to input high KV CT image data and low KV CT image data into a machine-learned network trained to predict bleeding regions given image data; the image processor is further configured to generate a visualization of the bleeding region predicted by the machine-learned network. A display is configured to display the visualization.

The present invention is defined by the following claims, and nothing in this section should be taken as a limitation on those claims. Further aspects and advantages of the invention are discussed below in conjunction with the preferred embodiments and may be claimed later, either individually or in combination.

Drawings

The components and figures are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. Moreover, in the figures, like reference numerals designate corresponding parts throughout the different views.

Fig. 1 depicts an example of a CT system.

Fig. 2 depicts a system for automatic detection and quantification for traumatic bleeding according to an embodiment.

Fig. 3 depicts a method for automatic detection of traumatic bleeding according to an embodiment.

Fig. 4 depicts an example dual energy CT system.

Fig. 5 depicts a method for automatic detection and quantification for traumatic bleeding according to an embodiment.

Fig. 6 depicts a system for automatic detection and quantification for traumatic bleeding according to another embodiment.

Detailed Description

Embodiments are provided for detecting and quantifying traumatic bleeding using a whole-body scan. A dual energy CT scan is performed on a patient to generate a plurality of imaging data sets. A plurality of imaging datasets are input into a machine-learned network trained to identify regions with bleeding based on different imaging datasets. The regions are automatically quantized and ranked. An overall risk assessment is provided to the operator.

Fig. 1 depicts an example CT imaging system 100. A target 110 (e.g., a patient) is positioned on a table 120, which table 120 is configured via a motorized system to move the table to a plurality of positions through a circular opening 130 in the CT imaging system 100. The X-ray source 140 (or other radiation source) and the detector element(s) 150 are part of a CT imaging system and are configured to rotate around the subject 110 on the gantry while the subject is inside the opening 130. The rotation may be combined with movement of the couch to scan along the longitudinal extent of the patient. Alternatively, the gantry moves the source 140 and detector 150 in a helical path around the patient. In CT imaging system 100, a single rotation may take about one second or less. During rotation of the X-ray source 140 and/or the detector, the X-ray source 140 generates a narrow fan (or cone) beam of X-rays that passes through a target portion of the body of the subject 110 being imaged. Detector element(s) 150 (e.g., multi-ring detector elements) are opposite X-ray source 140 and register X-rays that pass through the body of the object being imaged and, in the process, the recordings are used to create snapshots of the images. Many different snapshots at many angles through the object are acquired through one or more rotations of the X-ray source 140 and/or the detector element(s) 150. The image data generated from the acquired snapshots is transmitted to a control unit that stores the image data or processes the image data based on the snapshots into one or several cross-sectional images or volumes of the interior of the body (e.g., internal organs or tissues) of the subject scanned by CT imaging system 100.

While CT scanning is useful for certain medical diagnostics, standard CT scanning includes drawbacks for detecting traumatic bleeding. For identification and quantification, a general solution for the whole body is required, as the potential location of bleeding may be anywhere (e.g., brain, GI, thorax, abdominal cavity, etc.). Additionally, there is a great variability in the appearance of traumatic bleeding with respect to different zones and different types of injury. Finally, the bleeding area may include similar intensity values to other structures such as bone and blood vessels.