Abstract: A dynamic prototype convolution network (DPCN) can achieve sufficient information interaction between support features from a support image and query features from a query image for performing Few-shot semantic segmentation (FSS). A dynamic convolution module (DCM) can generate dynamic filters from a support foreground. Then, information interaction can be achieved by convolution operations over query features, such as by using these dynamic filters. A support activation module (SAM) and a feature filtering module (FFM) can be used to mine context information from a query feature. The SAM can learn to generate a pseudo mask for a query image. The FFM can refine the pseudo mask to filter background information from a query feature. Thus, information both from query and support can be used to achieve more accurate prediction. The DPCN can be used to perform k-shot segmentation.

Abstract: In a method and system for providing visual information about a tumour location in human or animal body, an electromagnetic tumour sensor is provided in the tumour and tracked to determine its location in space, which is mapped to a tumour model. A surgical tool sensor is provided on a surgical tool, and tracked to determine its location in space, which is mapped to a surgical tool model. The body is scanned to obtain information about an anatomical structure. A reference sensor is provided on the body, and tracked to determine its location in space, which is mapped to the anatomical structure. A virtual image is displayed showing the tumour model, located with the at least one tumour sensor, in spatial relationship to the surgical tool model, located with the at least one surgical tool sensor, and the anatomical structure, located with the at least one reference sensor.

Abstract: In a method and system for providing visual information about a tumour location in human or animal body, an electromagnetic tumour sensor is provided in the tumour and tracked to determine its location in space, which is mapped to a tumour model. A surgical tool sensor is provided on a surgical tool, and tracked to determine its location in space, which is mapped to a surgical tool model. The body is scanned to obtain information about an anatomical structure. A reference sensor is provided on the body, and tracked to determine its location in space, which is mapped to the anatomical structure. A virtual image is displayed showing the tumour model, located with the at least one tumour sensor, in spatial relationship to the surgical tool model, located with the at least one surgical tool sensor, and the anatomical structure, located with the at least one reference sensor.

Abstract: An example of sporadic motion that causes difficulty in CT scanning is gas pockets moving around the rectum. The invention allows the automatic detection of such movements, by enhancing low density features around the prostate in the individual X-ray images, projecting these features on the cranio-caudal axis (assuming that the gas predominantly moves in this direction) to form a 1-dimensional image, and combining successive ID projections to form a 2D image. Moving gas will produce tilted lines in this image, identifying an angular range that needs to be discarded. Such a process can be used in an image processing apparatus of a CT scanner.

Abstract: Methods and systems are provided for protecting a critical structure during the administration of radiation treatment to a patient. A register receives proposed positions for one or more radiation beams with respect to a critical structure. A processor predicts a cumulative dose volume for the critical structure based on the dose distribution, and determines if the cumulative dose volume exceeds a tolerance value. If the cumulative dose volume exceeds the tolerance value, the dose distribution may be translated at least in part based on a relationship between the cumulative dose volume and the dose distribution position.

Abstract: Methods and systems are provided for protecting a critical structure during the administration of radiation treatment to a patient. A register receives proposed positions for one or more radiation beams with respect to a critical structure. A processor predicts a cumulative dose volume for the critical structure based on the dose distribution, and determines if the cumulative dose volume exceeds a tolerance value. If the cumulative dose volume exceeds the tolerance value, the dose distribution may be translated at least in part based on a relationship between the cumulative dose volume and the dose distribution position.

Abstract: An in-line 4D cone beam CT reconstruction algorithm queues a limited number of projection images such that the phase determination algorithm can look-ahead. At regular intervals, the queue is scanned and those images which have enough look-ahead to obtain phase information are filtered and back-projected. The algorithm thus keeps up with the image acquisition speed and produces a 4D reconstruction within a few seconds of the end of scanning.

Abstract: The present invention provides a method and an apparatus for reconstructing images from data acquired during radiotherapy. The approach is based on summing the imaging data acquired while both therapeutic and imaging source are active, with that acquired when only the therapeutic source is active. Further correction of the summed data can lead to reconstructed images of excellent quality.

Abstract: An example of sporadic motion that causes difficulty in CT scanning is gas pockets moving around the rectum. The invention allows the automatic detection of such movements, by enhancing low density features around the prostate in the individual X-ray images, projecting these features on the cranio-caudal axis (assuming that the gas predominantly moves in this direction) to form a 1-dimensional image, and combining successive ID projections to form a 2D image. Moving gas will produce tilted lines in this image, identifying an angular range that needs to be discarded. Such a process can be used in an image processing apparatus of a CT scanner.

Abstract: Artifacts in the reconstructed volume data of cone beam CT systems can be removed by the application of respiration correlation techniques to the acquired projection images. To achieve this, the phase of the patients breathing is monitored while acquiring projection images continuously. On completion of the acquisition, projection images that have comparable breathing phases can be selected from the complete set, and these are used to reconstruct the volume data using similar techniques to those of conventional CT. Any phase can be selected and therefore the effect of breathing can be studied. It is also possible to use a feature in the projection images such as the patient's diaphragm to determine the breathing phase. This feature in the projection images can be used to control delivery of therapeutic radiation dependent on the patient's breathing cycle, to ensure that the tumor is in the correct position when the radiation is delivered.

Abstract: An apparatus, method and software module for selecting phase-correlated images from the output of a scanner such as a cone beam CT scanner operates by collapsing the images derived from the series from two dimensions to one dimension by summing the intensities of pixels along a dimension transverse to the one dimension, producing a further image from a composite of the one-dimensional images obtained from images in the series, analysing the further image for periodic patterns, and selecting from the series images having like phase in that periodic pattern. If desired, a plurality of reconstructions can be derived at different phases. The analysis of the further image for periodic patterns can include comparing the one-dimensional images therein, to identify a movement of features in that dimension. This allows (inter alia) the accurate determination of the breathing cycle in a patient and a concomitant improvement in the quality of CT scans by using phase-correlated images.

Abstract: Artifacts in the reconstructed volume data of cone beam CT systems can be removed by the application of respiration correlation techniques to the acquired projection images. To achieve this, the phase of the patients breathing is monitored while acquiring projection images continuously. On completion of the acquisition, projection images that have comparable breathing phases can be selected from the complete set, and these are used to reconstruct the volume data using similar techniques to those of conventional CT. Any phase can be selected and therefore the effect of breathing can be studied. It is also possible to use a feature in the projection images such as the patient's diaphragm to determine the breathing phase. This feature in the projection images can be used to control delivery of therapeutic radiation dependent on the patient's breathing cycle, to ensure that the tumor is in the correct position when the radiation is delivered.

Abstract: An apparatus, method and software module for selecting phase-correlated images from the output of a scanner such as a cone beam CT scanner operates by collapsing the images derived from the series from two dimensions to one dimension by summing the intensities of pixels along a dimension transverse to the one dimension, producing a further image from a composite of the one-dimensional images obtained from images in the series, analysing the further image for periodic patterns, and selecting from the series images having like phase in that periodic pattern. If desired, a plurality of reconstructions can be derived at different phases. The analysis of the further image for periodic patterns can include comparing the one-dimensional images therein, to identify a movement of features in that dimension. This allows (inter alia) the accurate determination of the breathing cycle in a patient and a concomitant improvement in the quality of CT scans by using phase-correlated images.

Abstract: Artefacts in the reconstructed volume data of cone beam CT systems can be removed by the application of respiration correlation techniques to the acquired projection images. To achieve this, the phase of the patients breathing is monitored while acquiring projection images continuously. On completion of the acquisition, projection images that have comparable breathing phases can be selected from the complete set, and these are used to reconstruct the volume data using similar techniques to those of conventional CT. Any phase can be selected and therefore the effect of breathing can be studied. It is also possible to use a feature in the projection image(s) such as the patient's diaphragm to determine the breathing phase. This feature in the projection images can be used to control delivery of therapeutic radiation dependent on the: patient's breathing cycle, to ensure that the tumour is in the correct position when the radiation is delivered.