Deep learning based image segmentation method including biodegradable stent in intravascular optical tomography image
Inventors
NAM, Hyeong-Soo • DOLGOV, Aleksandr • Lee, Sang-won • CHOI, Sogi • Yoo, Hong-Ki • Kim, Hyung-il
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11972574-B2
Publication Date
2024-04-30
Expiration Date
Abstract
Disclosed is an image segmentation device. The image segmentation device may include: a storage unit storing a segmentation model learned so as to segment at least one predetermined object; and at least one processor inputting input data into the segmentation model and segmenting at least one predetermined object in the input data, in which the segmentation model may include an encoder including at least one dimension reduction block reducing a dimension of the input data, a decoder including at least one dimension increase block increasing the dimension of output data output from the encoder by using data output from at least one dimension reduction block, and an auxiliary classification model receiving the output data and recognizing whether a specific object is included in the output data.
Core Innovation
The invention relates to an image segmentation device that segments predetermined objects in input data by using a non-transient, computer-readable storage unit storing a segmentation model and at least one processor. The segmentation model is trained using learning images so as to segment at least one predetermined object, and the processor inputs data into the segmentation model and segments the predetermined object in the input data. The predetermined object includes at least one of a strut and tissue of a biodegradable stent, and the input data is an intravascular optical coherence tomography (OCT) image.
The segmentation model includes an encoder with at least one dimension reduction block reducing a dimension of the input data, and a decoder with at least one dimension increase block increasing the dimension of output data outputted from the encoder using data outputted from at least one dimension reduction block. In addition, the segmentation model includes an auxiliary classification model receiving the output data outputted from the encoder and recognizing whether a specific object is included in the output data to produce a binary classification result. This auxiliary classification result is used during optimization so that a feature value for strut segmentation of the encoder and the decoder is further optimized when training the segmentation model.
During training, the processor optimizes weight values included in the decoder to express each of the predetermined objects included in the learning images in a final feature value map of the decoder, and optimizes weight values included in the encoder and the auxiliary classification model to recognize whether the specific object is included in the learning images to produce the binary classification result. Further optimization of the encoder and the decoder comprises minimizing a loss function L calculated using a term for the binary classification result of the auxiliary classification model. The learning images include a plurality of first learning images including a first biodegradable stent with a strut having a first thickness and a plurality of second learning images including a second biodegradable stent with a strut having a second thickness different from the first thickness.
Claims Coverage
The document includes at least one independent claim. The independent claim specifies an encoder-decoder segmentation architecture combined with an auxiliary binary classification model for object-presence recognition, optimized jointly during training, using intravascular OCT images and learning images of biodegradable stents with different strut thicknesses.
Encoder-decoder segmentation with dimension reduction and dimension increase blocks
The segmentation model includes an encoder including at least one dimension reduction block reducing a dimension of the input data, and a decoder including at least one dimension increase block increasing the dimension of output data outputted from the encoder by using data outputted from at least one dimension reduction block.
Auxiliary object-inclusion binary classification model
The segmentation model includes an auxiliary classification model receiving the output data outputted from the encoder and recognizing whether a specific object is included in the output data to produce a binary classification result.
Joint optimization using auxiliary binary classification loss to further optimize strut segmentation features
The processor optimizes weight values included in the encoder and the auxiliary classification model so as to recognize whether the specific object is included in the learning images to produce the binary classification result through the auxiliary classification model so that a feature value for strut segmentation of the encoder and the decoder is further optimized when training the segmentation model, wherein further optimization comprises minimizing a loss function L calculated using a term for the binary classification result of the auxiliary classification model.
Biodegradable stent strut and tissue segmentation from intravascular OCT
The at least one predetermined object includes at least one of a strut and tissue of a biodegradable stent, and the input data is an intravascular optical coherence tomography (OCT) image.
Training data spanning biodegradable stent strut thickness variations
The learning images include a plurality of first learning images including a first biodegradable stent configured by a strut having a first thickness, and a plurality of second learning images including a second biodegradable stent configured by a strut having a second thickness different from the first thickness.
The independent claim is centered on segmenting biodegradable stent strut and tissue from intravascular OCT using an encoder-decoder model, while jointly training the encoder and decoder with an auxiliary binary classification model that detects whether a specific object is included, using a loss function term tied to the auxiliary binary classification result and learning images that vary by strut thickness.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?