Method and system for intracerebral hemorrhage detection and segmentation based on a multi-task fully convolutional network
Embodiments of the disclosure provide systems and methods for detecting an intracerebral hemorrhage (ICH). The system includes a communication interface configured to receive a sequence of image slices and an end-to-end multi-task learning model. The sequence of image slices is the head scan images of a subject acquired by an image acquisition device. The end-to-end multi-task learning model includes an encoder, a bi-directional Convolutional Recurrent Neural Network (ConvRNN), a decoder, and a classifier. The system further includes at least one processor configured to extract feature maps from each image slice using the encoder, capture contextual information between adjacent image slices using the bi-directional ConvRNN, and detect the ICH of the subject using the classifier based on the extracted feature maps of the image slices and the contextual information or segment each image slice using the decoder to obtain an ICH region based on the extracted feature maps of the image slice.
1. A system for detecting an intracerebral hemorrhage (ICH), comprising:
a communication interface configured to receive a sequence of image slices and an end-to-end multi-task learning model, the sequence of image slices being head scan images of a subject acquired by an image acquisition device, the end-to-end multi-task learning model comprising an encoder, a bi-directional Convolutional Recurrent Neural Network (ConvRNN), a decoder, and a classifier; and
at least one processor, configured to:
extract feature maps from each image slice using the encoder;
capture contextual information between adjacent image slices using the bi-directional ConvRNN; and
detect the ICH of the subject using the classifier based on the extracted feature maps of the image slices and the contextual information or segment each image slice using the decoder to obtain an ICH region based on the extracted feature maps of the image slice.
2. The system of claim 1 , wherein the at least one processor is further configured to detect the ICH of the subject using the classifier and to segment each image slice using the decoder in parallel.
3. The system of claim 2 , wherein the at least one processor is further configured to estimate an ICH volume for the subject based on the ICH regions in the image slices.
4. The system of claim 1 , wherein to detect the ICH, the at least one processor is further configured to determine:
a slice-level ICH label for each image slice; or
a subject-level ICH label for the subject.
5. The system of claim 1 , wherein to detect the ICH, the at least one processor is further configured to determine:
a slice-level ICH subtype label for each image slice; or
a subject-level ICH subtype label for the subject.
6. The system of claim 1 , wherein the contextual information is determined based on a spatial relationship between every two adjacent image slices among the sequence of image slices.
7. The system of claim 1 , wherein the end-to-end multi-task learning model is a Fully Convolutional Network (FCN), wherein the encoder is a Convolutional Neural Network (CNN).
8. The system of claim 1 , wherein the bi-directional ConvRNN is a bi-directional Convolutional Long Short-Term Memory (ConvLSTM) or a bi-directional Convolutional Gated Recurrent Unit (ConvGRU).
9. The system of claim 1 , wherein the end-to-end multi-task learning model is trained using a joint loss function including a weight combination of a slice-level classification loss, a subject-level classification loss, a segmentation loss, and an ICH volume loss.
10. The system of claim 1 , wherein the end-to-end multi-task learning model further includes an attention module, and the at least one processor is further configured to enhance the extracted feature maps using the attention module.
11. A method for detecting an intracerebral hemorrhage (ICH), comprising:
receiving a sequence of image slices and an end-to-end multi-task learning model, the sequence of image slices being head scan images of a subject acquired by an image acquisition device, the end-to-end multi-task learning model comprising an encoder, a bi-directional Convolutional Recurrent Neural Network (ConvRNN), a decoder, and a classifier;
extracting, by at least one processor, feature maps from each image slice using the encoder;
capturing, by the at least one processor, contextual information between adjacent image slices using the bi-directional ConvRNN; and
detecting, by the at least one processor, the ICH of the subject using the classifier based on the extracted feature maps of the image slices and the contextual information or segmenting, by the at least one processor, each image slice using the decoder to obtain an ICH region based on the extracted feature maps of the image slice.
12. The method of claim 11 , wherein detecting the ICH of the subject using the classifier and segmenting each image slice using the decoder are performed in parallel.
13. The method of claim 12 , further comprising estimating an ICH volume for the subject based on the ICH regions in the image slices.
14. The method of claim 11 , wherein detecting the ICH further comprises determining:
a slice-level ICH label for each image slice; or
a subject-level ICH label for the subject.
15. The method of claim 11 , wherein detecting the ICH further comprises determining:
a slice-level ICH subtype label for each image slice; or
a subject-level ICH subtype label for the subject.
16. The method of claim 11 , wherein the contextual information is determined based on a spatial relationship between every two adjacent image slices among the sequence of image slices.
17. The method of claim 11 , wherein the end-to-end multi-task learning model is a Fully Convolutional Network (FCN), wherein the encoder is a Convolutional Neural Network (CNN).
18. The method of claim 11 , wherein the bi-directional ConvRNN is a bi-directional Convolutional Long Short-Term Memory (ConvLSTM) or a bi-directional Convolutional Gated Recurrent Unit (ConvGRU).
19. The method of claim 11 , wherein the end-to-end multi-task learning model is trained using a joint loss function including a weight combination of a slice-level classification loss, a subject-level classification loss, a segmentation loss, and an ICH volume loss.
20. A non-transitory computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by at least one processor, performs a method for detecting an intracerebral hemorrhage (ICH), the method comprising:
receiving a sequence of image slices and an end-to-end multi-task learning model, the sequence of image slices being head scan images of a subject acquired by an image acquisition device, the end-to-end multi-task learning model comprising an encoder, a bi-directional Convolutional Recurrent Neural Network (ConvRNN), a decoder, and a classifier;
extracting feature maps from each image slice using the encoder;
capturing contextual information between adjacent image slices using the bi-directional ConvRNN; and
detecting the ICH of the subject using the classifier based on the extracted feature maps of the image slices and the contextual information or segmenting each image slice using the decoder to obtain an ICH region based on the extracted feature maps of the image slice.