Varian Medical Systems International Ag

Siemens Healthineers International Ag

Hannu Laaksonen

Janne Iimari Nord

Esa Kuusela

Sami Petri Perttu

Maria Cordero Marcos

Jan Schreier

Elena Czeizler

María Cordero Marcos

Jarkko Yriana Peltola

Shahab Basiri

Mikko Oskari Hakala

Christopher Boylan

Ville Pietila

Tomi Ruokola

Perttu Niemela

Joakim Pyyry

Lauri Halko

Alexander Maslowski

María Isabel Cordero-Marcos

Perttu Niemelae

A control circuit accesses a plurality of information items that each correspond to a resultant dose volume histogram shape for a corresponding different radiation treatment plan. The control circuit then trains a machine learning model to predict a desired dose volume histogram shape using that plurality of information items as a training corpus.

Machine learning prediction of dose volume histogram shapes

A control circuit accesses a radiation treatment plan for a given patient. The control circuit then generates dose volume histogram information as a function of the radiation treatment plan and automatically assesses the dose volume histogram information to identify any anomalous results. Generating that information can comprise, at least in part and for example, generating at least one dose volume histogram curve. The latter may comprise generating at least one dose volume histogram curve for each of a plurality of different patient structures (such as one or more treatment volumes and/or one or more organs-at-risk).

Detecting anomalous dose volume histogram information

Example methods for radiotherapy treatment planning using deep learning engines are provided. One example method may comprise obtaining first image data associated with a patient; generating first feature data by processing the first image data associated with a first resolution level using a first processing pathway; generating second feature data by processing second image data associated with a second resolution level using a second processing pathway; and generating third feature data by processing third image data associated with a third resolution level using a third processing pathway. The example method may also comprise generating a first combined set of feature data associated with the second resolution level, and a second combined set of feature data associated with the first resolution level based on the first feature data and the first combined set. Further, the example method may comprise generating output data associated with radiotherapy treatment of the patient.

Methods and systems for radiotherapy treatment planning using deep learning engines

Example methods for adaptive radiotherapy treatment planning using deep learning engines are provided. One example method may comprise obtaining treatment image data associated with a first imaging modality and planning image data associated with a second imaging modality. The treatment image data may be acquired during a treatment phase of a patient. Also, planning image data associated with a second imaging modality may be acquired prior to the treatment phase to generate a treatment plan for the patient. The method may also comprise: in response to determination that an update of the treatment plan is required, processing, using the deep learning engine, the treatment image data and the planning image data to generate output data for updating the treatment plan.

Methods and systems for adaptive radiotherapy treatment planning using deep learning engines

Systems and methods for anatomical structure segmentation in medical images using multiple anatomical structures, instructions and segmentation models.

Systems and methods for automatic segmentation in medical imaging with multiple anatomical structure segmentation models

A radiation treatment plan three-dimensional dose prediction machine learning model is trained using a training corpus that includes a plurality of radiation treatment plans that are not specific to a particular patient and wherein the training corpus includes some, but not all, possible patient volumes of interest. Information regarding the patient (including information regarding at least one volume of interest for the patient that was not represented in the training corpus) is input to the radiation treatment plan three-dimensional dose prediction machine model. The latter generates predicted three-dimensional dose distributions that include a predicted three-dimensional dose distribution for the at least one volume of interest that was not represented in the training corpus.

Machine learning-based generation of 3D dose distributions for volumes not included in a training corpus

A control circuit accesses patient information including anatomical image information of the patient, segmentation information corresponding to the anatomical image information, and a dose map for the radiation treatment plan. The control circuit then generates at least one organ-specific three-dimensional risk map as a function of the patient information and presents that risk map to a user via a display.

Method and apparatus for radiation treatment plan risk assessment

Disclosed herein are methods and systems for predicting a virtual bolus in order to generate a radiation therapy treatment plan comprising training, by a processor, a machine-learning model using a training dataset comprising a set of medical images corresponding to a set of previously performed radiation therapy treatments, each medical image comprising at least one planning target volume and a non-anatomical region added to the medical image; and executing, by the processor, the machine-learning model using a medical image not included within the training dataset, the machine-learning model predicting an attribute of a non-anatomical region for the medical image not included in the training dataset.

Machine-learning modeling to generate virtual bolus attributes

Disclosed herein are systems and methods for training a machine learning model for automatic organ segmentation. A processor receives an image of one or more pre-contoured organs, the image comprising a plurality of voxels. The processor executes a machine learning model using the image to output predicted organ labels for the plurality of voxels of the image. The processor determines differences between corresponding predicted organ labels and expected organ labels for the plurality of voxels. The processor determines radiation dose levels that correspond to the plurality of voxels of the image. The processor determines weights for the plurality of voxels based on the radiation dose levels of the respective voxels. The processor then trains the machine learning model based on the differences and the weights for the plurality of voxels.

Using radiation dose information for automatic organ segmentation model training

One or more medical images of a patient are processed by a first neural network model to determine a region-of-interest (ROI) or a cut-off plane. Information from the first neural network model is used to crop the medical images, which serves as input to a second neural network model. The second neural network model processes the cropped medical images to determine contours of anatomical structures in the medical images of the patient. Each of the first and second neural network models are deep neural network models. By use of cropped images in the training and inference phases of the second neural network model, contours are produced with sharp edges or flat surfaces.

Systems and methods for image cropping and anatomical structure segmentation in medical imaging

A control circuit accesses historical information regarding previously optimized radiation treatment plans for different patients and processes that information to determine the relative importance of different clinical goals. The circuit then facilitates development of a particular plan for a particular patient as a function of the relative importance of the clinical goals. By one approach the control circuit can be configured as a radiation treatment plan recommendation resource that accesses a database of radiation treatment plan formulation content items including at least one of a radiation treatment plan template, an auto-planning algorithm, and an auto-segmentation algorithm. By one approach the control circuit can be configured to, when presenting automatically-generated radiation treatment plans to a user, also co-present an opportunity for the user to signal to a remote entity that none of the plans are acceptable and that the user will instead employ a user-generated plan for the particular patient.

Methods and apparatus pertaining to radiation treatment plans

A treatment planning apparatus includes: a modeler configured to obtain a model definition, wherein the model definition comprises a first quality metric of a first clinical goal; and a treatment planner having: a model trainer configured to obtain a set of existing treatment plans following desired clinical practice, and to perform model training to obtain a trained model based on the existing treatment plans and the first quality metric of the first clinical goal; an objective generator configured to generate a cost function based on the trained model; and an optimizer configured to determine a treatment plan based on the cost function.

Growing community of inventors

Espoo, Finland

Hannu Laaksonen