Search

EP-3968277-B1 - SEGMENTATION DEVICE

EP3968277B1EP 3968277 B1EP3968277 B1EP 3968277B1EP-3968277-B1

Inventors

  • ARAI, YOSHINORI
  • NISHIMURA, YUU
  • YOSHIKAWA, HIDEKI
  • SADAKANE, TOMOYUKI

Dates

Publication Date
20260506
Application Date
20201006

Claims (12)

  1. A segmentation device (3) comprising: an input unit (32) configured to receive an input of data of a constituent maxillofacial region which is a maxillofacial region or a partial region of a maxillofacial part, the constituent maxillofacial region including a first, biologically important region and a second region that is located outside the biologically important region; a calculation unit (34) configured to perform segmentation of the biologically important region and a region of interest in the second region using the data of the constituent maxillofacial region input to the input unit (32) and a previously generated learning model, and to calculate a three-dimensional position of the biologically important region and three-dimensional positional relationship information between the biologically important region and the region of interest; and an output unit (38) configured to output information based on a result of calculation from the calculation unit (34), wherein the learning model is a learning model which is generated using training data such that segmentation data of the biologically important region is output when the data of the constituent maxillofacial region is input and training data such that segmentation data of the region of interest is additionally output when the data of the constituent maxillofacial region is input, wherein the data of the constituent maxillofacial region is image data which is acquired by an X-ray CT scan or an MRI scan of the constituent maxillofacial region, wherein the biologically important region is at least one region of blood vessels, neural tubes, and a mandibular canal passing through the constituent maxillofacial region and a biological tissue passing through the mandibular canal, wherein the calculation unit (34) is configured to calculate a distance between the region of interest and the biologically important region as an inter-element distance, wherein the calculation unit (34) is configured to additionally generate an image of the second region, wherein the output unit (38) is configured to present a synthetic image of the image of the second region generated by the calculation unit (34) and an image of the segmentation data of the biologically important region, wherein the input unit (32) is configured to receive an input of a user's operation of designating a treatment target position, wherein the output unit (38) is configured to present an indicator of the treatment target position in an image of the constituent maxillofacial region in accordance with the designated treatment target position, and wherein the calculation unit (34) is configured to calculate a positional relationship between the biologically important region and the indicator of the treatment target position.
  2. The segmentation device (3) according to claim 1, wherein the output unit (38) is configured to display information of the inter-element distance calculated by the calculation unit (34).
  3. The segmentation device (3) according to claim 1 or 2, wherein the learning model includes: a first learning model which is generated using first training data such that segmentation data of a tooth region is output when the data of the constituent maxillofacial region is input; and a second learning model which is generated using second training data such that segmentation data of the biologically important region is output when the data of the constituent maxillofacial region and the segmentation data of the tooth region are input, and wherein the calculation unit (34) is configured to acquire the segmentation data of the tooth region using the data of the constituent maxillofacial region input to the input unit and the first learning model, and to perform segmentation of the biologically important region using the acquired segmentation data of the tooth region, the data of the constituent maxillofacial region input to the input unit, and the second learning model.
  4. The segmentation device (3) according to claim 3, wherein the learning model includes a third learning model which is generated using third training data such that the inter-element distance is output when the segmentation data of a tooth region and the segmentation data of the biologically important region are input, and wherein the calculation unit (34) is configured to calculate the inter-element distance using the segmentation data of the tooth region, the segmentation data of the biologically important region, and the third learning model.
  5. The segmentation device (3) according to claim 4, wherein the calculation unit (34) is configured to calculate a difficulty level of an implant treatment or a difficulty level of tooth extraction in accordance with the inter-element distance.
  6. The segmentation device (3) according to any one of claims 1 to 5, wherein information of an implant is additionally input to the input unit (32), and wherein the calculation unit (34) is configured to calculate a distance between the implant and the biologically important region when the implant is implanted on the basis of the information of the implant input to the input unit (32).
  7. The segmentation device (3) according to claim 5, wherein the output unit (38) is configured to present information of an implant which is usable on the basis of the result of calculation of the difficulty level.
  8. The segmentation device (3) according to any one of claims 1 to 7, wherein the input unit is configured to receive an operation of moving the indicator.
  9. The segmentation device (3) according to any one of claims 1 to 8, wherein the indicator of the treatment target position is an indicator of an implant.
  10. The segmentation device (3) according to claim 5 or 7, wherein the calculation unit (34) is configured to calculate a distance between the indicator of the treatment target position and the biologically important region as an inter-element distance.
  11. The segmentation device (3) according to claim 10, wherein the output unit (38) is configured to issue an alarm when the indicator of the treatment target position is close to the biologically important region by less than the inter-element distance or overlaps the biologically important region.
  12. The segmentation device (3) according to claim 11, wherein the output unit (38) is configured to present information of an implant which is usable on the basis of the result of the calculation of the difficulty level.

Description

TECHNICAL FIELD The present disclosure relates to a segmentation device. BACKGROUND Technology of performing segmentation on an image or the like obtained by an X-ray CT scan (for example, see Patent Document 1 (Japanese Unexamined Patent Publication No. H8-215192) or EP 3 449 421 B1) is known. SUMMARY In the related art, segmentation of a biological tissue in a medical image has been mathematically performed on the basis of CT values, concentration values, or the like. In this case, there is a problem in that it is difficult to segment tissues with close CT values, concentration values, or the like. A person's intervention (determination) is required for segmentation in consideration of an influence of conditions at the time of imaging or variables such as individual differences. Accordingly, there is demand for improvement in segmentation accuracy without requiring a person's intervention. An objective of the present disclosure is to provide a segmentation device that can improve segmentation accuracy. This is achieved by the claimed subject-matter. According to an aspect of the present disclosure, there is provided a segmentation device including: an input unit configured to receive an input of data of a maxillofacial region or a constituent maxillofacial region which is a partial region of a maxillofacial part; a calculation unit configured to perform segmentation of a biologically important region using the data of the constituent maxillofacial region input to the input unit and a previously generated learning model, and to calculate a three-dimensional position of the biologically important region in the constituent maxillofacial region; and an output unit configured to output information based on a result of calculation from the calculation unit. The learning model is a learning model which is generated using training data such that segmentation data of the biologically important region is output when the data of the constituent maxillofacial region is input. The data of the constituent maxillofacial region is image data which is acquired by an X-ray CT scan or an MRI scan of the constituent maxillofacial region. The biologically important region is at least one region of blood vessels, neural tubes, and a mandibular canal passing through the constituent maxillofacial region and a biological tissue passing through the mandibular canal. With this segmentation device, segmentation of a biologically important region is performed using a constituent maxillofacial region and a previously generated learning model. The learning model is a learning model which is generated using training data such that segmentation data of the biologically important region is output when the data of the constituent maxillofacial region is input. The data of the constituent maxillofacial region is image data which is acquired by an X-ray CT scan or an MRI scan of the constituent maxillofacial region. Accordingly, it is possible to segment a biologically important region from image data acquired by an X-ray CT scanner or an MRI scanner. By performing segmentation using the learning model in this way, a likelihood of improvement in segmentation accuracy increases, for example, in comparison with a case in which segmentation is mathematically performed on the basis of a CT value, a concentration value, or the like. With improvement in accuracy, a likelihood of a person's intervention not being required also increases. The learning model is a learning model which is generated using the training data such that segmentation data of a region of interest in a biologically normal region which is a region outside of the biologically important region in the constituent maxillofacial region is additionally output when the data of the constituent maxillofacial region is input, and the calculation unit is configured to perform segmentation of the region of interest. Accordingly, it is possible to segment the region of interest from the constituent maxillofacial region. The calculation unit is configured to calculate three-dimensional positional relationship information between the biologically important region and the region of interest. Accordingly, it is possible to understand a three-dimensional positional relationship between the biologically important region and the region of interest. The region of interest may be at least one region of a tooth region, a region which is occupied by an artifact implanted in the tooth region, a boundary region between a jawbone and the tooth region, a boundary region between the jawbone and the artifact, and an alveolar region, and the learning model may be a learning model which is generated using the training data such that segmentation data of each region of interest is output when the data of the constituent maxillofacial region is input. Accordingly, it is possible to segment each region of interest from the constituent maxillofacial region. The learning model may include: a first learning model w