Automated detection of lung conditions for monitoring thoracic patients undergoing external beam radiation therapy

Patent metadata

Field	Value
Publication number	US-12592309-B2
Application number	US-201917417406-A
Country	US
Kind code	B2
Filing date	Dec 16, 2019
Priority date	Dec 24, 2018
Publication date	Mar 31, 2026
Grant date	Mar 31, 2026

Abstract

A computerized system (SRS) for radiation therapy support. The system comprises an input interface (IN) for receiving an input image acquired by an imaging apparatus (IA 1 ). The input image represents a region of interest (ROI) internal of a patient (PAT) and acquired before delivery of a dose fraction by a radiation therapy delivery apparatus (RTD). A pre-trained machine learning unit (MLU) of the system is configured to process the input image to detect a medical condition. A communication component (RC) of the system is configured to provide, based on the detected medical condition, an indication for one or more clinical actions to be performed in relation to the patient.

First claim

The invention claimed is:

1 . A computerized system for radiation therapy support, comprising:
an input interface for receiving, during a therapy delivery phase, an input image comprising a 2D projection image acquired by an imaging apparatus, wherein the imaging apparatus is part of a radiation therapy delivery device, the input image representing at least a region of interest of a patient and acquired before delivery of a dose fraction, among multiple dose fractions associated with a treatment plan for the patient, by the radiation therapy delivery device; a pre-trained machine learning unit configured for processing the input image, based on a machine learning model pre-trained on training imagery, the processing including:
applying the input image to an input layer, the input layer comprising a matrix of nodes of a neural network, wherein a node of the matrix of nodes corresponds with a pixel of the input image; and
calculating an output vector, during the therapy delivery phase, the output vector including accumulated responses, wherein an index of entries of the output vector corresponds with a specified class, among a group of classes each representing a medical condition including a side effect of radiation therapy, wherein the specified class represents an indication of a presence of a respective medical condition;
a reporting component configured to provide, based on the output vector data and during the treatment delivery phase, an indication for one or more clinical actions to be performed in relation to the patient, wherein at least one of the one or more clinical actions as indicated by the reporting component includes an adaptation of the treatment plan.

2 . The system of claim 1 , wherein the one or more clinical actions include re-imaging the patient using a different imaging apparatus or modality.

3 . The system of claim 1 , wherein the reporting component comprises a communications interface, wherein the providing of the indication for the one or more clinical actions by the reporting component includes the communications interface transmitting a message through a communication network to a recipient.

4 . The system of claim 1 , including a visualizer configured to form an enriched image including the input image and a graphical rendering of a location in the input image in relation to the medical condition represented by the specified class.

5 . The system of claim 1 , wherein the machine learning unit is arranged in a neural network architecture.

6 . The system of claim 1 , comprising a correlator component configured to correlate patient data and the medical condition represented by the specified class with a treatment outcome value, indicative of a probability of a treatment outcome to be expected in relation to the patient.

7 . The system of claim 1 , comprising a knowledge builder component configured to populate, into a data repository, patient data in association with the medical condition represented by the specified class.

8 . An arrangement, comprising the system of claim 1 and the radiation therapy delivery device.

9 . A computer-implemented method for radiation therapy support, the method comprising:
receiving, during a therapy delivery phase, an input image comprising a 2D projection image acquired by an imaging apparatus, wherein the imaging apparatus is part of a radiation therapy delivery device, the input image representing at least a region of interest of a patient and acquired before delivery of a dose fraction, among multiple dose fractions associated with a treatment plan for the patient, by the radiation therapy delivery device; processing, during the therapy delivery phase, to detect a medical condition including a side effect caused by radiation therapy treatment, the processing based on a machine learning model pre-trained on training imagery, the processing comprising:
applying the input image to an input layer, the input layer comprising a matrix of nodes of a neural network, wherein a node of the matrix of nodes corresponds with a pixel of the input image; and
calculating an output vector, during the therapy delivery phase, the output vector including accumulated responses, wherein an index of entries of the output vector corresponds with a specified class, among a group of classes each representing a medical condition including a side effect of radiation therapy, wherein the specified class represents an indication of a presence of a respective medical condition;
providing the output vector including the indication of a presence of the respective medical condition, and providing during the therapy delivery phase, based on the output vector, an indication for one or more clinical actions to be performed in relation to the patient, wherein at least one of the one or more clinical actions includes an adaptation of the treatment plan.

10 . A computer program element, which, when being executed by at least one processing unit, is adapted to cause the processing unit to perform the method of claim 9 .

11 . A non-transitory computer readable medium having stored thereon the program element of claim 10 .

12 . The method of claim 9 , wherein the one or more clinical actions include re-imaging the patient using a different imaging apparatus or modality.

13 . The method of claim 9 , wherein the one or more clinical actions includes transmitting a message through a communication network to a recipient.

14 . The method of claim 9 , including forming an enriched image including the input image and graphically rendering of a location in the input image in relation to the medical condition represented by the specified class.

15 . The method of claim 9 , wherein the machine learning model is arranged in a neural network architecture.

16 . The method of claim 9 , comprising correlating patient data and the medical condition represented by the specified class with a treatment outcome value, indicative of a probability of a treatment outcome to be expected in relation to the patient.

17 . The method of claim 9 , comprising populating, into a data repository, patient data in association with the medical condition represented by the specified class.

18 . The method of claim 9 , comprising augmenting the input image, including at least one of superimposing or overlaying a class activation map (CAM) on the input image.

19 . The method of claim 18 , wherein the CAM identifies image pixels in the input image having contributed greater than a specified threshold to the respective medical condition represented by the specified class.

Assignees

• Elekta Inc

Inventors

• Bondar Maria Luiza
• Kustra Jacek Lukasz
• Consoli Sergio

Classifications

• A61N2005/1074

  Details of the control system, e.g. user interfaces · CPC title

• A61N5/1039

  using functional images, e.g. PET or MRI · CPC title

• G06N20/00

  Machine learning · CPC title

• G16H30/40

  for processing medical images, e.g. editing · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title