Methods and systems for predicting patient reported outcome after radiotherapy using functional imaging and radiation doses

Abstract

An apparatus comprises a processor and memory that stores instructions to cause the processor to receive user data that includes images of a body and organ. The processor is further caused to generate, via computed tomography imaging, a first image map and a second image map based on the user data. The processor is further caused to analyze, via spatial normalization of each image, the first image map and the second image map to determine strength of correlation of each tissue point of each image. The processor is further caused to execute a trained machine-learning model to generate a dose distribution image map using the first image map and the second image map as inputs. The processor is further caused to predict a change in a patient reported outcome for a user associated with the user data based on the dose distribution image map.

Description

[0001] METHODS AND SYSTEMS FOR PREDICTING PATIENT REPORTED OUTCOME AFTER RADIOTHERAPY USING FUNCTIONAL IMAGING AND RADIATION DOSES

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims the benefit of priority to U.S. Provisional Patent Application Serial No. 63 / 529,552 entitled " METHODS AND SYSTEMS FOR PREDICTING PATIENT REPORTED OUTCOME CHANGE AFTER RADIOTHERAPY USING FUNCTIONAL IMAGING AND RADIATION DOSES," filed July 28, 2023, the disclosure of which is incorporated herein by reference in its entirety.

[0004] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0005] This invention was made with government support under grant number R01CA200817 awarded by the National Institute of Health / National Cancer Institute. The government has certain rights in the invention.

[0006] BACKGROUND OF THE INVENTION

[0007] Cancer patients treated with radiotherapy suffer side effects due to treatment and can have reduced quality of life. An important way to measure side effect from treatment is patient reported outcomes (PROs). Measuring quality of life through patient reported outcomes is of importance because PRO data can be imperative to manage survivorship issues, be of prognostic value, help interpret clinical trials, and identify potential interventions. There are currently no robust methods to predict which patients will suffer clinically significant decline in quality of life as measured by PROs.

[0008] SUMMARY OF THE INVENTION

[0009] In one or more embodiments, a non-transitory, processor-readable medium storing instructions that when executed by a processor, can cause the processor to receive user data that includes a plurality of images of a body and organ. The processor can further be caused to generate, via computed tomography imaging, a first image map and a second image map based on the user data. The processor can be further caused to analyze, via spatial normalization of each image from the plurality of images of the user data, the first image map and the second image map to determine strength of correlation of a tissue point from a plurality of tissue points of each image. The processor can be further caused to execute a trained machine-learning model or mechanistic model to generate a functional image map using the first image map and the second image map as inputs. The functional image map can include a contour of ventilation of the organ based on a predetermined threshold for ventilation. The processor can be further be combined with a dose distribution and be caused to predict a change in a patient reported outcome for a user associated with the user data based on the dose distribution image map.

[0010] In one or more embodiments, an apparatus comprises a processor and a memory operatively coupled to the processor. The memory stores instructions that can cause the processor to receive user data that includes a plurality of images of a body and organ. The processor can be further caused to generate, via computed tomography imaging, a first image map and a second image map based on the user dat. The processor can be further caused to analyze, via spatial normalization of each image from the plurality of images of the user data, the first image map and the second image map to determine strength of correlation of a tissue point from a plurality of tissue points of each image. The processor can be further caused to execute mechanistic model or a trained machine-learning model to generate a functional image map using the first image map and the second image map as inputs. The functional image map can include a contour of ventilation of the organ based on a predetermined threshold for ventilation. The processor can be further be combined with a dose distribution and caused to predict a change in a patient reported outcome for a user associated with the user data based on the dose distribution image map.

[0011] In one or more embodiments, a method can include combining dose and functional imaging. The method can further include providing the combined dose and functional imaging information to a computer system including an artificial intelligence network. The artificial intelligence network can be trained using a plurality of patient information including dose and functional imaging from previous treatments and associated patient reported outcomes. The method can further include predicting a patient reported outcome for a target patient based on the computer system. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] For a fuller understanding of the nature and desired objects of the present invention, reference is made to the following detailed description taken in conjunction with the accompanying drawing figures wherein like reference characters denote corresponding parts throughout the several views.

[0013] FIG. 1 is a schematic illustration of a system for predicting patient reported outcome change, according to some embodiments.

[0014] FIG. 2 is an example flow diagram of a method for predicting patient reported outcome change, according to some embodiments.

[0015] FIG. 3 is another example flow diagram of a method for predicting patient reported outcome change, according to some embodiments.

[0016] FIGS. 4A-4B are diagrammatic illustrations depicting a concept of combining local function and local dose to predict for patient reported outcome changes following radiotherapy, according to some embodiments.

[0017] FIG. 5 is an example illustration of a patient reported outcome (PRO) information, according to some embodiments.

[0018] FIG. 6 are illustrative representations of images using 4-dimensional computed tomography, according to some embodiments.

[0019] FIG. 7 is are illustrative representations of images depicting 4-dimensional computed tomography ventilation and perfusion, according to some embodiments.

[0020] FIG. 8 is a graphical illustration depicting discrepancies between predicted and true postoperative pulmonary function test metrics, according to some embodiments.

[0021] FIG. 9 is a schematic diagram of a clinical trial dataset, according to some embodiments.

[0022] FIG. 10 is a schematic diagram of a Specific Aim (SA) evaluations and comparisons, according to some embodiments.

[0023] FIG. 11 are illustrative representations depicting 4-dimensional computed tomography for pre-surgery and post-surgery, according to some embodiments.

[0024] FIG. 12 is an illustrative representation depicting lung function imaging technique, according to some embodiments.

[0025] FIG. 13 is an illustrative representation depicting 4-dimensional computed tomography functional avoidance plans, according to some embodiments. DEFINITIONS

[0026] It is to be noted that any one or more of the aspects and embodiments described herein can be conveniently implemented using one or more machines (e.g., one or more compute devices that are utilized as a user compute device for an electronic document, one or more server devices, such as a document server, etc.) programmed according to the teachings of the present specification. Appropriate software coding can readily be prepared by skilled programmers based on the teachings of the present disclosure. Aspects and implementations discussed above employing software and / or software modules can also include appropriate hardware for assisting in the implementation of the machine executable instructions of the software and / or software module.

[0027] Such software can be a computer program product that employs a machine-readable storage medium. A machine-readable storage medium can be any medium that is capable of storing and / or encoding a sequence of instructions for execution by a machine (e.g., a compute device) and that causes the machine to perform any one of the methodologies and / or embodiments described herein. Examples of a machine-readable storage medium include, but are not limited to, a magnetic disk, an optical disc (e.g., CD, CD-R, DVD, DVD-R, etc.), a magnetooptical disk, a read-only memory "ROM" device, a random-access memory "RAM" device, a magnetic card, an optical card, a solid-state memory device, an EPROM, an EEPROM, and any combinations thereof. A machine-readable medium, as used herein, is intended to include a single medium as well as a collection of physically separate media, such as, for example, a collection of compact discs or one or more hard disk drives in combination with a computer memory. As used herein, a machine-readable storage medium does not include transitory forms of signal transmission.

[0028] Such software can also include information (e.g., data) carried as a data signal on a data carrier, such as a carrier wave. For example, machine-executable information can be included as a data-carrying signal embodied in a data carrier in which the signal encodes a sequence of instruction, or portion thereof, for execution by a machine (e.g., a compute device) and any related information (e.g., data structures and data) that causes the machine to perform any one of the methodologies and / or embodiments described herein. Examples of a compute device include, but are not limited to, an electronic book reading device, a computer workstation, a terminal computer, a server computer, a handheld device (e.g., a tablet computer, a smartphone, etc.), a web appliance, a network router, a network switch, a network bridge, any machine capable of executing a sequence of instructions that specify an action to be taken by that machine, and any combinations thereof. In one example, a compute device can include and / or be included in a kiosk.

[0029] As used herein, the singular form “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.

[0030] Unless specifically stated or obvious from context, as used herein, the term “about” is understood as within a range of normal tolerance in the art, for example within 2 standard deviations of the mean. “About” can be understood as within 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, 0.5%, 0.1%, 0.05%, or 0.01% of the stated value. Unless otherwise clear from context, all numerical values provided herein are modified by the term about.

[0031] As used in the specification and claims, the terms “comprises,” “comprising,” “containing,” “having,” and the like can have the meaning ascribed to them in U.S. patent law and can mean “includes,” “including,” and the like.

[0032] Unless specifically stated or obvious from context, the term “or,” as used herein, is understood to be inclusive.

[0033] Ranges provided herein are understood to be shorthand for all of the values within the range. For example, a range of 1 to 50 is understood to include any number, combination of numbers, or sub-range from the group consisting 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, or 50 (as well as fractions thereof unless the context clearly dictates otherwise).

[0034] DETAILED DESCRIPTION OF THE INVENTION

[0035] The disclosure described herein relates to devices and associated methods for predicting patient reported outcomes.

[0036] Proposed herein are methods to innovatively combine local radiation dose and local organ function imaging and devise dose / function-based methods to predict for changes in PROs following radiotherapy. In some embodiments, methods described herein can include a novel functional imaging technique or process to predict patient reported outcome surgical side effects for cancer patients. Provided herein are schematic examples demonstrating combing local dose and function to predict for PROs for lung cancer patients. The novel method and system will generate organ functional information at every voxel. The novel method and system will combine organ functional information at every voxel with dose information at every voxel (referred to as ‘dose-function,’ metrics). The novel method will define dose-function metrics for entire organs and will be used to predict for changes in patient reported outcomes after patients are treated with cancer therapies.

[0037] The novel method and system will enable early prediction for which patients are most likely to suffer clinically significant decline as determined by PROs. The novel method and system will enable a clinical decision support tool for early intervention for patients most likely to suffer clinically significant decline as determined by PROs. The novel method and system will enable radiation and surgical planning interventional strategies to reduce risk of clinically significant decline as measured by PROs.

[0038] FIG. 1 is a schematic illustration of a system 100 for predicting patient reported outcome (PRO) change, according to some embodiments. The compute device 100 includes a processor 102 and a memory 103 that communicate with each other, and with other components, via a bus 105. The compute device 100 can also include a database 105. The bus 105 can include any of several types of bus structures including, but not limited to, a memory bus, a memory controller, a peripheral bus, a local bus, and any combinations thereof, using any of a variety of bus architectures. The compute device 100 can be or include, for example, a computer workstation, a terminal computer, a server computer, a handheld device (e.g., a tablet computer, a smartphone, etc.), any machine capable of executing a sequence of instructions that specify an action to be taken by that machine, and any combinations thereof. The compute device 100 can also include multiple compute devices that can be used to implement a specially configured set of instructions for causing one or more of the compute devices to perform any one or more of the aspects and / or methodologies described herein.

[0039] The compute device 100 can include a network interface 106 and I / O (input / output) interfaces 107. The network interface 106, can be utilized for connecting the compute device 100 to one or more of a variety of networks and one or more remote devices connected thereto. In other words, although not shown in FIG. 1, the various devices including computer device 101 can communicate with other devices via a network(s). The network (not shown in FIG. 1) can include, for example, private network, a Virtual Private Network (VPN), a Multiprotocol Label Switching (MPLS) circuit, the Internet, an intranet, a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a worldwide interoperability for microwave access network (WiMAX®), an optical fiber (or fiber optic)-based network, a Bluetooth® network, a virtual network, and / or any combination thereof. In some instances, the network can be a wireless network such as, for example, a Wi-Fi or wireless local area network (“WLAN”), a wireless wide area network (“WWAN”), and / or a cellular network. In other instances, the network can be a wired network such as, for example, an Ethernet network, a digital subscription line (“DSL”) network, a broadband network, and / or a fiber-optic network. In some instances, the compute device 100 can use Application Programming Interfaces (APIs) and / or data interchange formats (e.g., Representational State Transfer (REST), JavaScript Object Notation (JSON), Extensible Markup Language (XML), Simple Object Access Protocol (SOAP), and / or Java Message Service (JMS)). The communications sent via the network can be encrypted or unencrypted. In some instances, the network can include multiple networks or subnetworks operatively coupled to one another by, for example, network bridges, routers, switches, gateways and / or the like. The I / O interfaces 107 of the compute device 100 can be any suitable component(s) that enable communication between internal components of the compute device 100 and external devices.

[0040] The processor 102 can be or include, for example, a hardware based integrated circuit (IC), or any other suitable processing device configured to run and / or execute a set of instructions or code. For example, the processor 102 can be a general-purpose processor, a central processing unit (CPU), an accelerated processing unit (APU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a programmable logic array (PLA), a complex programmable logic device (CPLD), a programmable logic controller (PLC) and / or the like. In some implementations, the processor 102 can be configured to run any of the methods and / or portions of methods discussed herein.

[0041] The memory 103 can be or include, for example, a random-access memory (RAM), a memory buffer, a hard drive, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), and / or the like. In some instances, the memory can store, for example, one or more software programs and / or code that can include instructions to cause the processor 102 to perform one or more processes, functions, and / or the like. In some implementations, the memory 103 can include extendable storage units that can be added and used incrementally. In some implementations, the memory 103 can be a portable memory (e.g., a flash drive, a portable hard disk, and / or the like) that can be operatively coupled to the processor 102. In some instances, the memory 103 can be remotely operatively coupled with a compute device (not shown); for example, a remote database device can serve as a memory and be operatively coupled to the compute device. The memory 103 can include various components (e.g., machine- readable media) including, but not limited to, a random-access memory component, a read only component, and any combinations thereof. In one example, a basic input / output system (BIOS), including basic routines that help to transfer information between components within the compute device 100, such as during start-up, can be stored in memory 103. The memory 103 can further include any number of program modules including, for example, an operating system, one or more application programs, other program modules, program data, and any combinations thereof.

[0042] The memory can include, for example, user data 110, a first image map 111, a second image map 112, a machine-learning or mechanistic model 113, a training set 114, a dose distribution image map 115, outcome parameters 116, and / or patient-reported outcomes (PROs) 117. The user data 114 can include images (e.g., 3D images) of a body of a mammal (e.g., human) and an organ (e.g., lung, brain, heart, etc.). In some implementations, the user data 110 can include 3D computed tomography (CT) images captured over various phases of a respiratory cycle for the body and / or the organ. The images can provide temporal dimension that captures motion of the organ. The images of the user 101 can also include images during an end inhalation and an exhalation phase (e.g., fully inhale, fully exhaled, etc.). The images can also depict various intermediate phases of the respiratory cycle. In some implementations, the user data 110 can also include patient-specific data associated with a user. For instance, the user data 110 can include a PRO filled out by the user that includes information such as, for example, voice of the user, quality of life, social well-being, psychological concerns, and / or the like. In some cases, the user data 110 can also be referred to herein as “4-dimensional computed typography (4DCT) data.”

[0043] The first image map 111 and the second image map 112 can be generated via 4DCT- ventilation and 4DCT-perfusion imaging techniques. The first image map 111 can be, for example, a 4DCT-ventilation image map. In some implementations, the first image map 11 1 can include a visualization of lung ventilation generated based on changes in lung volume and density throughout different phases of the respiratory cycle. The first image map 111 can provide information about how air moves through the lungs during the breathing cycle, aiding in various clinical applications such as radiation therapy planning, disease assessment, and surgical planning.

[0044] The second image map 112 can be, for example, a 4DCT-perfusion image map. In some implementations, the second image map 112 can include a visualization of blood flow (perfusion) within tissues in the organ (e.g., lung). By analyzing how blood flows through the tissues over time, the second image map 112 can provide information for diagnosing and treating various conditions, including tumors, strokes, pulmonary embolisms, chronic lung diseases, and / or the like.

[0045] The database 106 can store information generated by the processor 102 and / or received at the processor 102. In some implementations, the database 106 can include, for example, hard disk drives (HDDs), solid-state drives (SSDs), USB flash drives, memory cards, optical discs such as CDs and DVDs, and / or the like. In some implementations, the database 106 can include a database (e.g., a cloud database, a local database, etc.) that can be different from the memory 103. For example, the memory 103 can be volatile, meaning that its contents can be lost when the compute device 100 is turned off. The database 106 can be configured to be persistent, meaning that its contents can be retained even when the compute device 100 is turned off. In some implementations, the database 106 can be configured to organize and manage large amounts of data, whereas the memory 103 can be configured to be used for temporary storage of data and program instructions. In some implementations, the database 106 can be configured to provide efficient and reliable storage and retrieval of data and can include features such as, for example, indexing, querying, and transaction management, while the memory 103 can be configured for rapid access and manipulation of data. In some implementations, the database 106 can include a collection of PROs, 4DCT data of multiple users, 4DCT-ventilation image maps, 4DCT-perfusion maps, and / or the like.

[0046] The machine-learning or mechanistic model 113 can include a set of model parameters such as weights, biases, or activation functions that can be executed to annotate and / or classify 4DCT data, 4DCT-ventilation images, 4DCT-perfusion images, , dose distribution image maps, PRO or PRO changes, and / or the like. The machine-learning model 1 13 can be executed during a training phase and / or an execution phase.

[0047] In the training phase, the machine-learning model 113 receives training data and optimizes (or improves) the set of model parameters of the machine-learning model 113. The set of model parameters are optimized (or improved) such that, for example, dose distribution image maps and / or PROs in the training data can be annotated and / or classified correctly with a certain likelihood of correctness (e.g., a pre-set likelihood of correctness). The training data can include 4DCT data, 4DCT-ventilation images, 4DCT-perfusion images, , dose distribution image maps, PRO or PRO changes, and / or the like.

[0048] In some instances, the training data can be divided into batches of data based on a memory size, a memory type, a processor type, and / or the like. In some instances, the resume document images can be divided into batches of data based on a type of the processor 104 (e.g., CPU, GPU, and / or the like), number of cores of the processor, and / or the like.

[0049] In some instances, the training data can be divided into the training set 114, a test set, and / or a validation set. For example, the training data can be randomly divided so that 60% of the training data is in the training set, 80% of the training data is in the test set, and 90% of the training data is in the validation set. The machine-learning model 113 can be iteratively optimized (or improved) based on the training set while being tested on the test set to avoid overfitting and / or underfitting of the training set. Once the machine-learning model 113 is trained based on the training set and the test set, a performance of the machine-learning model 113 can be further verified based on the validation set. In some implementations, the machinelearning model 113 can be trained using the training set 114 that includes a pre-treatment image correlated to a post-treatment dose distribution image map. In some cases, the training set 114 can also include a pre-treatment PRO of a user correlated to a post-treatment PRO of a user. In some cases, the training set 114 can also include 3D image scans (e.g., 4DCT images) of a user correlated to a parameters associated with a change in that user’s PRO.

[0050] In the execution phase, the machine-learning model 113 (that is trained in the training phase) receives the first image map 111 and / or the second image map 112 not among the set of image maps used in the training phase) and annotates and / or classifies them to produce the dose distribution image map 115. Because the execution phase is performed using the set model parameters that were already optimized during the training phase, the execution phase is computationally quick.

[0051] The machine-learning model 113 can be or include at least one of a supervised machinelearning model, an unsupervised machine-learning model, a deep neural network model (DNN), an artificial neural network (ANN) model, a fully connected neural network, a convolutional neural network (CNN), a residual network model, a region proposal network (RPN) model, a feature pyramid network (FPN) model, a generative adversarial network (GAN), a K-Nearest Neighbors (KNN) model, a Support Vector Machine (SVM), a decision tree, a random forest, an analysis of variation (ANOVA), boosting, a Naive Bayes classifier, and / or the like.

[0052] The dose distribution image map 115 can include, for example, a 3D, voxel -wise image of radiation dose distribution.. In some implementations, the outcome parameters 106 can include characteristics that are predictive for a user’s PRO 117. For instance, the outcome parameters can include, for example, specific radiation dose.

[0053] The memory 103 can include instructions that when executed by the processor 102, cause the processor 102 to receive the user data that 110 includes images of a body and organ of a user such as, for example, a lung. The memory 103 can further include instructions that when executed cause the processor 102 to generate, via computed tomography imaging (e.g., 4DCT- ventilation, 4DCT-perfusion, ec.), the first image map 111 and the second image map 112 based on the user data 110. In some implementations, generating the first image map 111 and the second image 112 can include instructions to further cause the processor 102 to segment the images of the organ to identify the organ while excluding pulmonary vasculatures and airways of the body during an inhale phase and an exhale phase of the user data, map corresponding tissue points on the organ in each image of the user data 110 during the inhale phase and the exhale phase, and calculate ventilation in each tissue point from the plurality of tissue points in each image to generate the first image map 111 and the second image map 112. In some cases, mapping corresponding tissue points can include capturing changes in ventilation and perfusion throughout a breathing cycle.

[0054] The memory 103 can further include instructions that when executed further cause the processor 102 to analyze, via spatial normalization of each image from the user data 110, the first image map 111 and the second image map 112 to determine strength of correlation of a tissue point from a plurality of tissue points of each image. In some implementations, the processor 102 can generate a 3D map that can be overlaid on each of the first image map 111 and the second image map 112 based on the analysis to provide information for potential treatment plan (e.g., radiation dose). In some cases, the analysis can include a voxel-based lung-function analysis.

[0055] The memory 103 can further include instructions that when executed further cause the processor 102 to execute the machine-learning model 113 to generate he dose distribution image map 115 using the first image map 111 and the second image map 112 as inputs. In some cases, the dose distribution image map 115 can include a contour of ventilation of the organ based on a predetermined threshold for ventilation (e.g., 15% or greater ventilation). The memory 103 can further include instructions to cause the processor 102 to predict a change the PRO 117 for a user associated with the user data 110 based on the dose distribution image map 115 and / or pretreatment images, PROs, etc.

[0056] FIG. 2 is an example flow diagram of a method 200 for predicting PRO change, according to some embodiments. At 205, the method 200 can include receiving user data that includes a plurality of images of a body and organ.

[0057] At 210, the method 200 can include generating, via CT imaging, a first image map (e.g., 4DCT-ventilation image map) and a second image map (4DCT-perfusion image map) based on the user data. In some cases, the CT imaging can include, for example, 4DCT-ventilation and 4DCT perfusion imaging. In some implementations, generating the first image map and the second image map can include segmenting the plurality of images of the organ to identify the organ while excluding pulmonary vasculatures and airways of the body during an inhale phase and an exhale phase of the user data, mapping corresponding tissue points on the organ in each image from the plurality of images of the user data in phase during the inhale phase and the exhale phase, and calculating ventilation in each tissue point from the plurality of tissue points in each image from the plurality of images to generate the first image map and the second image map. In some implementations, mapping corresponding tissue points can include capturing changes in ventilation and perfusion throughout a breathing cycle.

[0058] At 215, the method 200 can include analyzing, via spatial normalization of each image from the plurality of images of the user data, the first image map and the second image map to determine strength of correlation of a tissue point from a plurality of tissue points of each image. In some implementations,. The analysis can be executed via machine-learning. At 220, the method 200 can include executing a trained machine-learning model to generate a dose distribution image map using the first image map and the second image map as inputs, the dose distribution image map includes a contour of ventilation of the organ based on a predetermined threshold for ventilation (e.g., 15% or greater ventilation). The machine-learning model can include a supervised machine-learning model or unsupervised machine-learning model. In some cases, the machine-learning model can be trained using a training set that includes a pre-treatment image correlated to a post-treatment dose distribution image map

[0059] At 225, the method 200 can include predicting a change in a patient reported outcome for a user associated with the user data based on the dose distribution image map.

[0060] FIG. 3 is another example flow diagram of a method 300 for PRO change, according to some embodiments. At 305, the method 300 can include combining dose and functional imaging. At 310, the method 300 can include providing the combined dose and functional imaging information to a computer system including an artificial intelligence network, the artificial intelligence network having been trained using a plurality of patient information including dose and functional imaging from previous treatments and associated patient reported outcomes. At 315, the method 300 can include predicting a patient reported outcome for a target patient based on the computer system.

[0061] FIGS. 4A-4B are diagrammatic illustrations depicting a concept of combining local function and local dose to predict for PRO changes following radiotherapy, according to some embodiments. Schematic showing concept of combining local function and local dose to predict for patient reported outcome changes following radiotherapy. In some implementations, methods described herein can allow for predictive dose-response modeling of PROs for patients treated on functional avoidance trial.

[0062] FIG. 5 is an example illustration of a PRO information, according to some embodiments. The innovation of the methods described herein is that it combines functional imaging and predictive dose-response modeling to predict for PROs. In some implementations the novel lung function imaging technique as described herein can be developed using 4DCTs and image processing to generate lung ventilation maps (e.g., first image map 111 and second image map 112 of FIG. 1). In some cases, lung ventilation maps can also be referred to herein as “4DCT- ventilation / perfusion,” 4DCT-ventilation / perfusion images,” or “4DCT-ventilation / perfusion image maps.” In some implementations, 4DCT-ventilation functional avoidance can use 4DCT- images to generate treatment plans that avoid functional regions of an organ (e g., lung) aiming to reduce pulmonary toxicity. PROs provided by users / patients can be essential measure of quality -of-life following radiotherapy as they show that 1) they represent patient quality of life as reported by the patients themselves and 2) the patient perspective and the physician assessment of the patient perspective may not always be in accord. In some embodiments, the impact of the methods described herein can identify dose / dose-function metrics that can be used to minimize the probability that patients experience clinically significant decline, as measured by patients themselves.

[0063] In some implementations, users and patients can initially be enrolled on a 4DCT- ventilation functional avoidance trial in which they were administered the Functional Assessment of Cancer Therapy Lung (FACT-L) questionnaire at pre-treatment and 3 months post-treatment. The questionnaire can be configured to gather data on physical, social, emotional, functional, and pulmonary well-being. Based on validated methods, the FACT-TOI (Trial Outcome Index) and the FACT-LCS (Lung Cancer Subscale) metrics can be generated and percentage of clinically meaningful PRO declines can be calculated. The ability of standard organ-at-risk (lung, cord, esophagus, heart) dose and lung dose-function metrics (metrics that combine dose and function based on 4DCT-ventilation imaging) to predict for PRO clinically significant decline can be evaluated using tests and logistic regression. To further train a machine-learning model to predict improved and / or accurate changes in PRO and / or dose distribution. In some cases, normal lung doses, esophagus doses, and dose-function metrics can be considered to reduce the probably of patient reported clinically significant decline.

[0064] FIGS. 14-15 depict illustrative representations of lung function imaging technique and 4DCT functional avoidance plans, according to some embodiments. As shown in FIG. 14, 4DCT data and images can be used to produce 4DCT-ventilation image maps that show regions of ventilation. As shown in FIG. 15, 4DCT images can also be used to generate treatment plans avoiding functional regions of lung to reduce pulmonary toxicity.

[0065] The following examples further illustrate aspects of the present invention. However, they are in no way a limitation of the teachings or disclosure of the present invention as set forth herein.

[0066] EXAMPLES EXAMPLE-1 Prospective Trial of Functional Lung Avoidance Radiation Therapy for Lung Cancer: Quality of Life Report

[0067] Purpose. A novel form of lung function imaging can be described herein that uses 4DCT data to generate lung ventilation images (4DCT-ventilation). Functional avoidance uses 4DCT- ventilation to reduce doses to functional lung with the aim of reducing pulmonary side effects. A phase 2, multicenter 4DCT-ventilation functional avoidance clinical trial was completed. The purpose of this work was to quantify changes in patient-reported outcomes (PROs) for patients treated with functional avoidance and determine which metrics are predictive of PRO changes.

[0068] Materials and Methods'. Patients with locally advanced lung cancer receiving curativeintent radiation therapy were accrued. Each patient had a 4DCT-ventilation image generated using 4DCT data and image processing. PRO instruments included the Functional Assessment of Cancer Therapy-Lung (FACT-L) questionnaire administered pretreatment; at the end of treatment; and at 3, 6, and 12 months posttreatment. Using the FACT-Trial Outcome Index and the FACT-Lung Cancer Subscale results, the percentage of clinically meaningful declines (CMDs) were determined. A linear mixed-effects model was used to determine which patient, clinical, dose, and dose-function metrics were predictive of PRO decline.

[0069] Results. Of the 59 patients who completed baseline PRO surveys. 83% had non-small cell lung cancer, with 75% having stage 3 disease. The median dose was 60 Gy in 30 fractions. CMD FACT-Trial Outcome Index decline was 46.3%, 38.5%, and 26.8%, at 3, 6, and 12 months, respectively. CMD FACT-Lung Cancer Subscale decline was 33.3%, 33.3%, and 29.3%, at 3, 6, and 12 months, respectively. Although an increase in most dose and dose-function parameters was associated with a modest decline in PROs, none of the results were significant (all P > .053).

[0070] Conclusions'. The current work presents an innovative combination of use of functional avoidance and PRO assessment and is the first report of PROs for patients treated with prospective 4DCT-ventilation functional avoidance. Approximately 30% of patients had clinically significant decline in PROs at 12 months posttreatment.

[0071] Introduction

[0072] Patients with lung cancer frequently deal with ongoing symptoms and morbidity associated with poor quality of life (QOL).l Lung cancer treatment often includes chemotherapy, radiation, immunotherapy (IO), and surgery, which may all affect QOL. Radiation treatment is traditionally designed to limit lung doses, including mean lung dose and volume of lung that receives >20 Gy (lung V20 Gy).2 One limitation to the standard dose-volume approach is that it assumes that lung function is homogenous and that dose to all parts of the lung is equally damaging. Studies have shown that the majority of lung cancer patients have heterogeneous lung tissue often related to the tumor and comorbidities, such as chronic obstructive pulmonary disease (COPD).

[0073] Functional avoidance radiation is a technique developed to take advantage of the heterogeneous lung function to reduce pulmonary side effects. Functional avoidance attempts to limit doses to functional parts of a patient’s lung as defined by imaging based on the hypothesis that reduced doses to functional regions of the lung results in reduced pulmonary toxicity. 4DCT along with image processing techniques can generate a surrogate for lung ventilation that is referred to as 4DCT-ventilation. As 4DCT is routinely performed as part of the radiation treatment planning process; therefore, lung ventilation images can be generated from 4DCT data without burdening the patient with an extra imaging procedure. 4DCT-ventilation has been compared with other forms of functional lung imaging, including nuclear medicine planar ventilation-perfusion scans; single-photon emission computed tomography imaging; and more experimental forms of functional lung imaging, with promising global correlation results.

[0074] Modeling work has shown the feasibility of employing 4DCT-ventilation functional avoidance to reduce doses to functional portions of the lung without sacrificing tumor dose or exceeding neighboring normal tissue constraints. Based on data from retrospective modeling studies, a 2-institution functional avoidance clinical trial using 4DCT-ventilation was completed. The trial reported the primary endpoint of clinical toxicity and found that functional avoidance radiation therapy reduced the rate of grade 2 or higher radiation pneumonitis compared with the historical rate (14.9% with 4DCT-ventilation functional avoidance compared with 25% historical control).

[0075] Assessment of PROs is an important way to measure QOL, including any changes experienced after cancer treatment. Studies have shown that patients often view QOL after cancer treatment as the second most important factor when making decisions regarding cancer therapy options, with curing the cancer being the most important factor. Maintaining a high level of QOL after treatment for lung cancer is becoming increasingly important as therapies such as IO are being developed and patients are continuing to live longer. After completing treatment for lung cancer, patients may experience continued symptoms, including fatigue, cough, and dyspnea, that affect their ability to perform their desired tasks and chronically affect their health- related QOL. Recently published data suggest that patient self-reported QOL metrics may differ from clinician-scored toxicity and in some instances may be more reliable and accurate. Taken together, these studies underscore the importance of patient-reported outcome data when evaluating lung cancer therapies.

[0076] The hypothesis of 4DCT-ventilation functional avoidance is that pulmonary toxicity can be reduced, and QOL can be improved. Based on the assumptions that QOL is an important outcome metric when evaluating novel lung cancer therapies and that 4DCT-ventilation functional avoidance can affect QOL, PRO data were collected for patients treated in a 4DCT- ventilation functional avoidance clinical trial. The purpose of this work is to characterize pre- to posttreatment PRO changes for patients treated with 4DCT-ventilation functional avoidance and determine which patient, clinical, dose, and dose-function metrics are predictive of PRO changes.

[0077] Materials and Methods

[0078] Study Design

[0079] Patients were consented and enrolled on a multi-institution institutional review board- approved phase 2 clinical trial (ClinicalTrials.gov Identifier: NCT02528942). Patients were enrolled at 2 institutions: University of Colorado (IRB #14- 1856) and Beaumont Health System (IRB# 2016-037). Patients 18 and older were required to have a confirmed diagnosis of either non-small cell (NSCLC) or small cell lung cancer (SCLC), and be planned for curative chemoradiation. Patients were excluded from the study if they were planned for treatment with stereotactic body radiation or palliative doses of radiation, defined as <45 Gy. Chemotherapy and / or IO were allowed per physician discretion. There were no trial exclusion criteria based on baseline patient performance status or pulmonary function test (PFT) results.

[0080] The trial included a 4DCT-ventilation image heterogeneity inclusion criteria. On 4DCT- ventilation imaging, a 15% reduction in regional lung function near the tumor, as well as a qualitative noted ventilation defect determined as “yes” or “no” by the treating physician, were required. The rationale for the image heterogeneity criteria was that if their lung function was homogenous, patients would likely not benefit from sparing certain areas. The process of identifying a ventilation defect has been described previously in nuclear medicine.

[0081] Previously described image processing techniques along with 4DCT data were used to generate 4DCT-ventilation images for each patient. Briefly, the first step was to perform advanced lung segmentation (with pulmonary vasculature and airways excluded) in the end inspiration and end expiration phases of the 4DCT. Deformable image registration was then used to register lung voxels in each breathing phase. A density-change — based equation was then applied to calculate ventilation in each voxel. An example of a 4DCT-ventilation image for one enrolled patient is shown in FIG. 6. FIG. 6 depicts an image 600 obtained using 4DCT- ventilation for a patient enrolled in the trial. The 4DCT-ventilation image is normalized as a percentile image with brighter colors as denoted in, for example, region 604 representing functional portions of the lung and darker colors denoted in, for example region 602 showing ventilation defect regions. Use of 4DCT-ventilation allows radiation planning that preferentially spares the higher functioning parts of the lung using treatment planning techniques. Image 600 also shows planning target volume as denoted by region 606 and accompanying dose distribution with the 4DCT-ventilation image.

[0082] After 4DCT-ventilation functional imaging was created, a contour of the higher functioning portions of the lung was generated using a threshold of 15% or greater ventilation based on previously described data. The functional contour along with the radiation oncologists’ identification of gross tumor volume, planning target volume (PTV), and relevant organs at risk (OARs) were used for treatment planning. All treatment plans were generated using intensity modulated radiation therapy (IMRT) techniques and favorable arc geometry and optimization techniques to reduce doses to functional regions of the lung as defined by the functional contour. Specifically, arc start and stop angles were altered to reduce dose to the functional contour and the doses decreased to the functional contour using IMRT optimization objectives. No set functional objectives were used, as the objectives were at the discretion of the treatment planner.

[0083] The planner did not sacrifice PTV and standard OAR metrics in favor of reducing doses to functional portions of the lung. Not sacrificing PTV and standard OAR metrics in favor of reducing doses to functional portions of the lung ensured that the functional avoidance treatment plans would meet all standard thoracic radiation therapy criteria. An example of a functional avoidance treatment plan is shown in FIG. 6. Treatment planning constraints were in line with the recommendations and guidelines of the Quantitative Analysis of Normal Tissue Effects in the Clinic reports Radiation Therapy Oncology Group (RTOG) 0617, NRG LU 001, and RTOG 0538. Tumor coverage and OAR dose constraints were not compromised to reduce doses to the functional contour to ensure that the plans met standard lung cancer radiation plan evaluation criteria. Dosimetric PTV and OAR results for the clinical trial have been previously presented.

[0084] Health-related Quality of Life Measurement

[0085] The 4DCT-ventilation functional avoidance trial used a phase 2 design with a secondary endpoint of QOL. All patients had the opportunity to complete the PRO assessments. QOL was measured using 3 validated instruments: Functional Assessment of Cancer Therapy Lung (FACTLung), EuroQol 5 Dimension 5 Level (EQ-5D-L), and the visual analog scale (VAS).19- 22 QOL questionnaires were administered before treatment; at the end of treatment; and at 3, 6, and 12 months posttreatment.

[0086] The FACT-L questionnaire includes 36 items within 5 sections: Physical Well-Being (PWB), Social / Family Well-Being, Emotional Well-Being, Functional Well-Being (FWB), and the Lung Cancer Subscale (LCS). Questions are answered using a 5-point Likert scale ranging from 0 to 4 to evaluate the past 7 days. Subcomponents from the FACT-L questionnaire are then calculated using a weighted average of the responses to a specific set of questions. The 2 most frequently calculated subcomponents and the subcomponents used for the current work are the FACT Trial Outcome Index (FACT-TOI) and the FACT-Lung Cancer Subscale (FACT-LCS). The FACT- TOI is comprised of the PWB, FWB, and the LCS portions of the FACT-L questionnaire. The FACT-LCS consists of 7 items related to trouble breathing, chest tightness, and cough. A clinically meaningful change has been previously established as a change of >5 points in the FACT-TOI subcomponent and >2 points in the LCS subcomponent. The VAS ranges from 0 to 100, with 100 being the best health the patient can imagine. Validated clinically meaningful decline (CMD) in VAS is considered to be a change in 7 points.22 EQ-5D-L is a tool with 5 questions in each of the following domains; mobility, self-care, usual activities, pain, and anxiety / depression. The scores are converted to a health index ranging from 0 to 1, with the maximum score of 1 representing ideal health.

[0087] Patient-reported Outcome Analysis The differences in the FACT-TOI and FACT-LCS from pretreatment to each timepoint (end of treatment and at 3, 6, and 12 months posttreatment) were calculated for each patient. The difference was then classified as clinically meaningful or not based on a >5 point change for the FACT-TOI or >2 point change for the FACT-LCS. The mean of the change in scores (reported as the mean ± SD) for all patients and percentage of patients with clinically meaningful decline are presented. The mean change in score was calculated by first determining the difference in the FACT-TOI or FACT-LCS score from the time point of interest to baseline before determining the average of the change over the entire population. To evaluate longitudinal PRO change using methods previously presented, a linear mixed effects model was used to examine mean change in the FACT-TOI from baseline and mean change in FACTLCS from baseline at all post-treatment measurement times controlling for pretreatment values and accounting for the within-subject correlations among the same PROs measured at different time points. The average change and associated 95% Cis at each time point were computed.

[0088] PROs provide a critical clinical endpoint for functional avoidance radiation therapy. No studies have evaluated which clinical, patient, or treatment planning factors in the context of functional avoidance may be predictive of QOL and PRO changes. This study evaluated which patient, clinical, dose, and dose-function metrics were predictive of clinically significant decline in the FACT-TOI and FACT-LCS. Patient and clinical factors included sex, smoking status, performance status, age, whether the patient had COPD, histology, surgical status (scored as yes or no), and IO status (scored as yes or no). Standard dose metrics evaluated included mean lung dose, lung V20 Gy (volume received >20Gy), heart V40 Gy, and heart V60 Gy.

[0089] Dose-function metrics are metrics that combine dose and 4DCT-ventilation based function to determine whether doses are deposited to functional portions of the lung. Dose function metrics evaluated included structure-based threshold methods and were chosen based on previous work. To evaluate the effects of patient, clinical, dose, and dose-function metrics on PROs, linear mixed effect modeling was used. Baseline patient, clinical, dose, and dose-function variables were added to the linear mixed effects model, and the slope parameters and their 95% Cis were computed to determine how they were associated with changes in the FACT-TOI and FACT-LCS. The time dependence of the slope parameters at each time point was examined. A cross-sectional sub-analysis was performed (presented in Appendix A) using Fisher’s exact test to evaluate the percentage of clinically significant FACT-TOI and FACT-LCS declines for patients who did and did not receive treatment with TO and patients who did and did not undergo surgery.

[0090] The baseline VAS score (mean ± SD) as well as the mean of the changes (mean ± SD) at each time point are reported. The health index was calculated and reported from the EQ- 5D-L score and converted using the available crosswalk index calculator. The baseline EQ-5D-L score (mean ± SD) as well as the mean of the changes (mean ± SD) at each time point are reported.

[0091] Results

[0092] The median age was 65 (range, 45-86 years). Sixty-one percent of patients were female and 39% male. The median Karnofsky performance status (KPS) was 90 (range, 60-100), with 54.2% having COPD and 91.5% being current or former smokers. Forty-six patients were diagnosed with NSCLC (83%) and 44 (75%) with stage 3 disease. Nine (15.3%) underwent surgery as part of their treatment, including 8 who received a lobectomy and 1 a pneumonectomy. The median dose was 60 Gy (range, 45 Gy-66 Gy) delivered in 30 fractions (range, 23-33 fractions). Fifty-two patients (88.1%) received concurrent chemotherapy, and 16 (27.1%) were treated with IO while they were enrolled in the study.

[0093] Of the 59 evaluable patients with baseline QOL evaluations, 58 (98%), 54 (92%), and 43 (73%) completed questionnaires at 3, 6, and 12 months posttreatment, respectively. The PRO results using the FACT-L instrument for patients treated with 4DCT-ventilation functional avoidance are presented in Table 2. The percentage of patients with clinically significant declines in FACT-TOI at the end of treatment and at 3, 6, and 12 months posttreatment was 65.5%, 46.3%, 38.5%, and 26.8%, respectively. The percentage of patients with clinically significant LCS decline was 55.2%, 33.3%, 33.3%, and 29.3% at end of treatment and at 3, 6, and 12 months posttreatment, respectively. When restricting the data to patients with NSCLC, the results are similar to those for the overall cohort, with 46.5%, 36.6%, and 25.7% of patients experiencing FACT-TOI-based clinically significant decline at 3, 6, and 12 months, respectfully.

[0094] The FACT-TOI results show a large statistically significant decrease from pretreatment values at the end of treatment (slope = -10.81, P < .001) and 3 months posttreatment (slope = - 3.28, P = .028), followed by a nonsignificant decrease at 6 months and a nonsignificant increase at 12 months (both P > .101). Similarly, FACT-LCS declined at the end of treatment (slope = - 1 .51, P = .004), with nonsignificant declines at 3 months posttreatment (slope = -0.28, P = .608) and nonsignificant increases at 6 and 12 months posttreatment (all P > 0537).

[0095] Because there was evidence that the associations between heart V40 Gy and the FACT- TOI as well as between age and the FACT-LCS depended on the follow-up duration, the association was calculated for each timepoint. Although an increase in most dose and dosefunction parameters was associated with a modest decline in PROs from pretreatment, none of the evaluated factors were significantly predictive of FACT-TOI or FACT-LCS changes.

[0096] Regarding patient and clinical factors, current smokers, patients with stage 3 disease (relative to stage 1 and 2), patients with SCLC, and patients who experienced > grade 2 pneumonitis had a noted decline in both the FACT-TOI and FACT-LCS, although the results were not statistically significant (all P > .053). Cross-section results using Fisher’s exact test for clinically significant decline in the FACT-TOI and FACTLCS for patients who received surgery or IO as part of their treatment are provided in Appendix A. Patients who underwent surgery had lower rates of FACT-TOI clinically meaningful decline, but none of the differences are significant (all P > 0.132). Patients who received IO had higher rates of clinically meaningful FACT-TOI decline, but none of the differences were significant (all P > .534).

[0097] The average VAS ± SD was 71.5 ± 19.4 at baseline, with mean changes of -2.6 ± 20.4, 2.8 ± 14.6, 1.6 ± 22.2, 3.5 ± 15.1 at the end of treatment at 3 months, 6 months, and 12 months posttreatment, respectively. The average health index score using the EQ5D questionnaire (mean ± SD) was 0.825 ± 0.13 at baseline, with mean changes of -0.048 ± 0.129, -0.014 ± 0.095, -0.032 ± 0.138, and -0.013 ± 0.120 at the end of treatment and at 3, 6, and 12 months posttreatment, respectfully.

[0098] Discussion

[0099] The presented work is the first study to present a comprehensive analysis of patient- reported outcomes for patients treated with 4DCT-ventilation functional avoidance. An imaging modality that provides functional information without burdening the patient with an extra imaging procedure, 4DCT-ventilation has been shown to reduce pneumonitis compared with a historical control. The current work innovatively combined 4DCT-ventilation and QOL data to investigate how 4DCT-ventilation functional avoidance may have affected QOL outcomes. Less than half of patients (44.4%) had a clinically meaningful decline in the FACTTOI 3 months after treatment, and at 12 months after treatment, the percentage of patients with clinically meaningful decline decreased to 26.8%. When the assessment focused on the FACT-LCS, 35% and 30% of patients had a clinically meaningful decline at 3 and 12 months, respectively. Assessment using the VAS showed a similar rate of clinically significant decline of approximately 20%. These data demonstrate encouraging patient-reported QOL outcomes, and when taken together, the PFT and pneumonitis results provide a strong rationale for evaluating 4DCT-ventilation in a phase 3 trial.

[0100] Although the study was not powered to compare PRO data between cohorts, its results provide a reference for PRO results for patients treated with standard thoracic radiation therapy. RTOG 0617 was a national phase 3 clinical trial comparing standard dose (60 Gy) to high dose radiation (74 Gy) with concurrent chemotherapy while also evaluating efficacy of cetuximab in patients with locally advanced NSCLC. As a secondary endpoint of RTOG 0617, PROs for patients treated with 60 Gy on the RTOG 0617 protocol were presented. At 12 months after radiation therapy, 47.8% of patients treated in RTOG 0617 experienced clinically significant decline, whereas 22.8% of patients treated with functional avoidance experienced clinically significant decline.

[0101] There are several noteworthy differences between the patients treated in RTOG 0617 and patients treated with functional avoidance, including treatment technique (IMRT vs 3D planning), inclusion and exclusion of IO, inclusion and exclusion of surgery, baseline performance status, and baseline PFT results.

[0102] A longitudinal trend that appeared in both the cross-sectional results and the linear mixed effects model accounting for the wi thin-subject correlation is a degradation in PROs at the end of radiation therapy before a gradual return to baseline at 12 months. The longitudinal data are in line with those from other lung cancer PRO studies showing reduced QOL after radiation therapy treatment before a gradual return to baseline QOL 1 year after treatment. Although many aspects affect PROs, potential hypotheses for the explanation of the pattern of reduced QOL immediately after treatment with a gradual return to baseline can be due to side-effects and disease cure. Certain side effects from radiation therapy can occur at the end of treatment that can be expected to dissipate 1 year from treatment. With regard to disease cure, patients with a worse prognosis may have dropped out of the study before the 1 year follow-up; therefore, the data collected at the 1 year timepoint may represent patients with improved disease prognosis. The results of correlation analysis indicated several patient and clinical factors were possible predictive factors of PRO decline, including current smokers (relative to nonsmokers), surgical status (Appendix A), patients with stage 3 disease (relative to stage 1 and 2), patients with SCLC, and patients who experienced >grade 2 pneumonitis (Tables 4 and 5). However, the correlations were not statistically significant, possibly because the study was underpowered to assess PROs. It should be noted that the correlation trends were in line with clinical knowledge. Patients with stage 3 disease (relative to stage 1 and 2) and patients with SCLC (relative to NSCLC) have a worse prognosis and are therefore more likely to have worse QOL. Patients who experienced >grade 2 pneumonitis are more likely to have reduced QOL due to their pulmonary symptoms. The correlation between receiving surgery and experiencing less clinical decline is likely linked to generally having a better health profile at baseline to be considered a candidate for surgery. Overall, these data showing a link between patient and clinical factors and PROs should be interpreted as hypothesis generating, as the data are likely underpowered.

[0103] This study is the first to evaluate the ability of dose and dose-function metrics to predict clinically significant PRO decline in the functional avoidance setting using a linear mixed effects model. Although an increase in some dose and dose-function parameters was associated with a decline in the FACT-TOI, the results were not statistically significant. Although there has not been a comparison published of dose-response modeling of PROs for standard fractionated lung treatments, higher lung dose-volume metrics for patients experiencing clinically significant PRO decline who were treated with lung stereotactic body radiation has been noted. The mixed doseresponse findings underscore the complex relationship between treatment planning factors, patient factors, and clinical factors, and PROs and should be considered when evaluating functional avoidance in a phase 3 setting.

[0104] In interpreting the data, it is imperative to consider the specific questions that make up the PRO instruments relative to the goal of functional avoidance. The goal of functional avoidance is to reduce the rate of pulmonary toxicity and improve QOL for patients with lung cancer treated with radiation therapy. The FACT-L is the most frequently used instrument in assessing QOL for lung cancer patients treated with radiation therapy. Although the FACT-L focuses on lung cancer and asks patients about shortness of breath and cough, it also assesses clear thinking, appetite, and weight loss, none of which is hypothesized to improve by avoiding radiation to functional lung. The FACT-TOI includes the PWD and FWB subscales in addition to the FACT-LCS.

[0105] These subscales assess facets that could have been affected by functional avoidance, including less bothersome side effects and ability to work and enjoy life. Alternatively, it is possible that the FACT-TOI scores were driven by improved overall survival with more modern treatments, such as IO. Although there are limitations in using PRO instruments to quantify QOL relative to pulmonary toxicity, the data provide a complete and quantitative profile of QOL after functional avoidance. The data support the hypothesis that 4DCT-ventilation functional avoidance results in QOL that is as good as or better than QOL compared with a historical control. The PRO results, the radiation pneumonitis data, and the PFT data all suggest a promising toxicity profile and QOL with 4DCT-ventilation functional avoidance and warrant further comparison in a randomized trial. The results provide seminal data that can be used to provide clinical and dosimetric guidance in functional avoidance trials designed to evaluate improvement in patient QOL.

[0106] Conclusion

[0107] Use of 4DCT-ventilation functional avoidance offers the opportunity to generate radiation therapy plans that reduce pulmonary side effects without requiring an extra imaging procedure. The current work is the first report of PROs for patients treated in a prospective 4DCT-ventilation functional avoidance clinical trial. The data show that 4DCT-ventilation functional avoidance results in approximately 20% to 40% of patients with clinically significant decline, with this percentage decreasing from 3 to 12 months posttreatment. Although an increase in most dose and dose function parameters was associated with a modest decline in PROs from pretreatment, none of the evaluated patient, clinical, dose, or dose-function factors were significantly predictive of FACT-TOI or FACT-LCS changes. The presented PRO results provide positive evidence with which to investigate 4DCT functional avoidance for locally advanced lung cancer in a phase 3 setting.

[0108] EXAMPLE 2 - Novel Functional Imaging to Predict Surgical Side Effects for Lung Cancer Patients

[0109] Significance Methods described herein proposes to improve how lung-cancer patients are evaluated for resectional procedures by assessing a novel imaging modality and advanced computational methods for evaluating lung function. Lung cancer is the second-most diagnosed cancer and the leading cause of cancer-related death worldwide. Surgical resection is a primary mode of definitive treatment for patients with lung cancer. The complication with resectional methods for lung cancer is that patients may lack sufficient lung function to tolerate surgery. Lung-cancer patients have compromised lung function from their tumor or other accompanying thoracic conditions, such as emphysema. Patients who have poor lung function prior to surgery are at increased risk of developing serious and life-threatening thoracic complications after resection. Therefore, as part of a standard pre-surgery work-up, surgeons evaluate a patient’s pre-surgical lung function and predict what the patient’s lung function will be after the resection. The patientspecific predicted lung function informs a clinical decision regarding whether the patient is offered or denied surgery. Current methods for surgical lung-function evaluation involve pulmonary function tests (PFTs) in combination with anatomic approximations. However, studies have shown that current lung-function evaluation methods can be inaccurate and can produce errors in the predicted, post-surgical lung function of over 20%10. Critically, post- surgical lung function prediction errors can affect patient outcomes. If a patient’s lung function is erroneously underestimated, the patient may be subjected to additional tests or, in the worst case, wrongly denied surgery. If lung function is overestimated, the patient will be at a higher risk of morbidity after surgery. For example, a recent study reported a surgical complication rate of 26% (CTCAE Grade 2+). Improved lung-function assessment can substantially enhance the quality of care for lung-cancer patients being considered for surgery Functional lung imaging has been demonstrated to improve prediction of tumor response and clinical outcomes for lung-cancer patients, and evaluation of pulmonary comorbidities. The cardiothoracic surgery discipline relies on global (PFT) assessments and has yet to employ local, voxel-based functional lung imaging. This project proposes novel functional imaging and computational tools that have great potential to provide more accurate pre-surgical lung-function assessment and superior prediction of post- surgical outcomes. The novel lung-function imaging methods proposed herein are known as 4DCT-ventilation and 4DCT-perfusion. 4DCT-ventilation / perfusion produces 3D, high- resolution maps of lung ventilation / perfusion and is an excellent way to visualize which portions of the lung are functional and used for breathing without the need for invasive contrast delivery. FIG. 7 is are illustrative representations of images 701 / 711 depicting 4DCT- ventilation / perfusion, according to some embodiments. FIG. 7 shows examples of 4DCT- ventilation image 701 and 4DCT-perfusion image 711, which can include bright colors (as denoted in region 705) representing functional portions of the lung and the dark tones (as denoted in region 703) displaying ventilation / perfusion defect areas. The patient presented in the top row of images 701 / 711 has two lung tumors, outlined in region 707, with a ventilation defect surrounding one and a perfusion defect surrounding the other. The bottom image of 701 shows a patient with a right-central tumor, outlined in region 709, that is occluding a central airway and is causing a ventilation defect in the entire right lung. 4DCT-ventilation / perfusion imaging has been successfully integrated into Radiation Oncology and shown to improve clinical outcome prediction for lung-cancer patients receiving radiation therapy. The high spatial resolution of 4DCT-ventilation / perfusion enables the application of 2 other advanced computational tools that have found efficacy in radiation oncology: (i) voxel-based analysis of spatial associations between image features and clinical outcomes and (ii) machine-learning (ML) prediction of images. The methods described herein propose to assess the ability of the 3 cutting-edge computational tools (i.e., 1. novel lung-function imaging, 2. voxel-based analysis, 3. ML image prediction) to provide more accurate pre-surgical lung-function assessment and superior prediction of post-surgical lung function and side effects.

[0110] Strengths / weaknesses of Rigor of Prior Research

[0111] The thoracic surgical discipline continues to rely on global (i.e., PFT) assessments. Although proof-of-concept work has evaluated planar ventilation-perfusion (V / Q) nuclear medicine scans and more experimental functional imaging techniques in the cardiothoracic surgery domain, there are 3 significant barriers that have prevented the widespread adoption of functional imaging in surgical evaluation: 1) the previously investigated imaging modalities have considerable image-quality and patient-based shortcomings, 2) post-treatment lung function imaging has never before been investigated, and 3) the breadth of clinical endpoints have not been considered. Planar V / Q scans and other forms of experimental lung function imaging have provided limited improvement in surgical evaluation because the scans provide 2D data, can suffer from serious image artifacts, and have patient-based disadvantages in a contrast requirement, long scan time, and experimental imaging that may not be widely available. 4DCT- ventilation / perfusion has significant image-quality improvements and patient-based advantages that can enable its clinical adoption. Prior work has investigated functional imaging prior to resection and has not collected functional imaging data after the resection. This study provides functional imaging before and after surgery, enabling us to study the local, spatial changes of lung function after surgery. Previous studies have evaluated the impact of pre-surgical imaging on PFT metrics. For a complete picture of lung function, other clinical information is needed.

[0112] Innovation

[0113] The innovation within this proposal lies in 1) the approaches investigated and 2) the data used. More exacting surgical evaluation is more important than ever as cutting-edge surgical techniques enable surgeons to offer early-stage lung cancer patients minimally invasive surgeries that include lobectomies and segementectomies. The methods described herein aim to transform pre-surgical evaluation by proposing three novel approaches: 1) innovative lung-function imaging, 2) voxel-based local analysis, and 3) ML image prediction. The three novel approaches have been academically adopted in Radiation Oncology and Radiology disciplines. The benefits of each approach is independent of the others with the ability to work synergistically. In some embodiments, the proposed novel approaches can be readily translatable to the surgical domain where they will improve how lung-cancer patients are evaluated for resectional procedures.

[0114] The proposed novel approaches can be tested using an unprecedented dataset that was prospectively collected as part of an innovative cardiothoracic clinical trial (NCT03426306). The trial is titled, ‘A Novel Lung Function Imaging Modality as a Preoperative Evaluation Tool (LIME),’ and collected pre- and post-surgical 4DCT-ventilation / perfusion imaging, PFTs, clinical outcomes, and PROs for lung-cancer patients being considered for resection. The dataset collected herein enables 4DCT-ventilation / perfusion assessment pre- and post-surgery as well as consideration of additional important endpoints including clinical toxicities and PROs. The comprehensive dataset is primed and ready for analysis, enabling us to definitively evaluate the accuracy of these novel approaches.

[0115] Approach

[0116] Background and Preliminary Data

[0117] Role of imaging in Surgical Evaluation of Lung Cancer Patients For many patients with lung cancer, surgery remains the primary treatment option. The complication with resectional methods for lung cancer is that patients may not have sufficient lung function to tolerate surgery. As part of a standard pre-surgery work-up, surgeons use PFT data to estimate every patient’s lung function. PFTs use spirometry to measure various airflow metrics. PFT scores are used to calculate percent predicted postoperative (%PPO) PFT values by scaling the pre-surgical PFT values by the amount of lung function that will be lost due to surgery. In the conventional workflow, the amount of lung function lost is estimated anatomically as the relative proportion of lung segments that will be removed, as in Equation 1: where PFT0is the pre-surgical PFT metric and 19 is the total number of lung segments. The anatomic approach assumes that lung function is evenly distributed among the lung segments. Several studies have attempted to address the limitations of conventional anatomic scaling with more advanced imaging techniques such as quantitative 3DCT, contrast-enhanced perfusion (Q) MRI, nuclear medicine planar V / Q scans, and single-photon emission computed tomography (SPECT) imaging. Critically, widespread clinical adoption of V / Q nuclear medicine and SPECT scans into the routine surgical evaluation workflow has been limited by poor spatial resolution and imaging artifacts that can cause large post-surgical prediction errors in the patient’s lung function. Erroneous lung-function assessment can have a clinically significant adverse impact on patient care. Furthermore, key patient-based disadvantages of Q-MRI, V / Q scans, and SPECT scans, including high imaging costs, long acquisition times, and the need for invasive contrast delivery, have precluded adoption of these techniques. 4DCT- ventilation / perfusion can address all the image-quality and patient-based limitations and enable the wide-spread adoption of functional imaging into pre-surgical lung function evaluation.

[0118] DCT-ventilation / perfusion Imaging

[0119] The novel lung-function assessment method proposed here was developed in Radiation Oncology and uses image processing to calculate lung ventilation and perfusion from standard, non-contrast, 4-Dimensional Computed Tomography (4DCT) imaging. 4DCTs (also known as gated CTs) are standard 3-dimensional (3D) CT images resolved into the various phases of the breathing cycle. 4DCTs provide a ‘movie’ of the moving lung anatomy and are routinely used in radiation oncology. Mathematical algorithms and methods have been developed to robustly calculate 4DCT-ventilation to calculate 4DCT-perfusion. 4DCT-ventilation / perfusion produces a full 3D map of ventilation / perfusion, has high imaging resolution, provides quantitatively meaningful voxel values, displays multi-modal data (anatomical and functional), and has a modest imaging dose of 3 cGy.

[0120] The image processing methods of 4DCT-ventilation and 4DCT-perfusion can be described below. The first step is to acquire a standard, non-contrast 4DCT. 4DCTs are a standard option on most radiation oncology and radiology CT scanners. The imaging technique as described herein can only require advanced image processing on non-contrast 4DCT data. The method uses the 4DCT data to map lung voxels from the inhale to exhale phase and approximates the 4DCT-ventilation and 4DCT-perfusion by capturing the changes in air content

[0121] (ventilation) and pulmonary blood mass (perfusion) throughout the breathing cycle. The mathematical formulation of 4DCT-ventilation / perfusion is represented by Equation 2: exhale density, *sinhale density, is volume change, and MC is mass change density. Equation 2 purports that the exhale density is a linear function of the inhale density with the slope being the volume change, which represents a surrogate for ventilation, and the y- intercept being the mass change density, which represents a surrogate for perfusion.

[0122] Clinical Integration of 4DCT- ventilation / perfusion

[0123] Although the proposed project as described in this example is in the field of thoracic surgery, progress in radiation oncology demonstrates that 4DCT-ventilation imaging has matured enough for clinical integration. Based on strong evidence from preliminary data, 4DCT- ventilation / perfusion can be incorporated into 2 clinical trials. The first was a multi-institutional prospective clinical trial (R01CA200817) that used 4DCT-ventilation to spare functional portions of the lung during radiation therapy. The results demonstrated that using 4DCT- ventilation in radiation treatment can reduce the rates of pulmonary toxicity for lung cancer patients. The second clinical trial evaluated the accuracy of novel 4DCT-perfusion methods. The pilot study (NCT03183063) compared 4DCT-perfusion and nuclear-medicine perfusion for 5 patients being evaluated for a pulmonary embolism. The trial data showed good agreement between nuclear-medicine perfusion and 4DCT-perfusion.

[0124] Demonstration of 4DCT-ventilation for Surgical Evaluation

[0125] The concept of using 4DCT-ventilation for resectional surgery evaluation was assessed using retrospective data. This study included 16 lung-cancer patients who were evaluated for both surgery and radiation therapy, and therefore had both PFTs and 4DCT imaging collected. The PFT data was used to calculate the standard metric used to evaluate surgical candidates: percent predicted postoperative (%PPO) PFTs based on both the 4DCT-ventilation images and conventional anatomic methods (Equation 1). This study revealed that there was good global agreement between novel 4DCT-ventilation and conventional approximation for about 60% of patients. For the other 40% of patients, there were substantial discrepancies between using 4DCT-ventilation and conventional approximation of %PPO PFT metrics; discrepancies that would result in different clinical decisions. It is also revealed that there are differences in lungfunction assessment using the two methods; there was no data to determine which method was more accurate in predicting true, post-surgical lung function.

[0126] Pilot Clinical Trial of Lung Functional Imaging as Preoperative Evaluation Tool

[0127] This study includes 4DCT-ventilation / perfusion imaging for 65 lung cancer patients were collected and considered for surgical resection. To estimate the accuracy of conventional %PPO PFT calculations, a preliminary analysis was performed of the 65 lung cancer patients from the pilot clinical trial, who will be included in the proposed study. The %PPO PFTs were calculated using Equation 1 for a collection of 4 common PFT measures for each patient and compared the conventional, anatomic-based %PPO PFT measures to the true postoperative PFT values. The results revealed that anatomic approximations of %PPO PFTs provided an error (root-mean- squared error) of 20% when compared to postoperative PFTs and tended to underestimate the true postoperative PFT by an average of 11%, as shown in FIG. 8. FIG. 8 is a graphical illustration depicting discrepancies between %PPO and true PFT test metrics depicted in a Bland-Altman plot, according to some embodiments. The average relative discrepancy is plotted in red and the dashed red lines indicate the 25th and 75th percentile error levels. The baseline correlation between conventional predictions and true postoperative PFTs was 0.74 (Pearson correlation coefficient; p = 1.49E-17). The proposed analysis described herein is designed to definitively answer the question of whether 4DCT-ventilation / perfusion can improve upon anatomic approximations to accurately predict post-surgical lung function.

[0128] Limitation of Conventional, Global Surgical Evaluation

[0129] The current paradigm of predicting post-surgical lung function relies on predicting global metrics. Conventional global metric prediction has 2 major limitations: 1) inability to capture spatially complex changes in lung function 2) lack of local, voxel-level information provided to clinicians. %PPO PFTs are calculated by scaling the pre-surgical PFT values by the amount of lung function that will be lost due to surgery. However, the global approach remains unable to capture the complexity of the surgical impact on local lung function. In some cases, local lung function may counterintuitively improve due to the removal of diseased tissue enabling healthy lung tissue to expand, likely contributing to conventional %PPO PFTs underestimating true postoperative PFTs by as much as 50-80% for individual patients. In other cases, lung segments close to the surgical resection exhibit decreased function post-surgery, despite their presumed preservation. Predicting complex changes in local lung function requires local, voxel-based prediction of post-surgical lung function.

[0130] Historically, the standard surgical procedure for managing lung cancer was a pneumonectomy (removal of an entire lung) or a lobectomy (removal of a lobe). Because of advances in early screening methods and surgical procedures, there has been a substantial recent trend toward performing minimally invasive surgeries that preserve lung volume, known as segmentectomies. Because segmentectomies remove smaller portions of the lung, they require local, voxel-based lung-function information to provide clinicians with a more accurate prediction of postoperative lung function.

[0131] Preliminary Data Summary

[0132] Lung-cancer patients who are considered for surgical resection undergo global lungfunction evaluation to determine whether they have sufficient lung function to withstand surgery. Some current evaluation methods can produce inaccurate lung-function assessments. Advanced imaging and computational methods show potential to deliver considerable image-quality and patient-based advantages over conventional lung-function evaluation techniques.

[0133] Methods and studies as described herein can demonstrate that in the radiation oncology setting that 4DCT-ventilation / perfusion, ML prediction, and voxel-based analysis methods are ready to be clinically integrated. The resulting dataset reveals that significant differences between conventionally predicted and true postoperative PFT measures persist. The next logical step is thorough analysis to evaluate whether advanced computational tools can more accurately predict post-surgical outcomes.

[0134] Prospective Clinical Trial Dataset

[0135] This project can leverage an unprecedented dataset that was prospectively collected as part of a pilot clinical trial (as shown in FIG. 11). FIG. 9 is a schematic diagram of a clinical trial dataset that measures PFT and PRO metrics, according to some embodiments. The multi- institutional clinical trial (NCT03426306) completed accrual in January 2023 with 65 patients who were considered for resectional surgery. As part of the standard surgical work up, all patients underwent PFTs. As part of the study, patients underwent pre-surgical 4DCT imaging and evaluations of clinical symptoms (surgical thoracic complications and patient 02 requirements) and patient-reported outcomes (PROs). Three months post-surgery, patients underwent a follow-up PFT, which will provide the true, ‘gold standard’ in the analysis of the predictive capability of these novel computational tools. In addition to follow-up PFTs, the 3- month, post-surgical evaluation collected 4DCTs, clinical endpoints, and PROs. Although the postoperative PFTs provide the ‘gold standard’, clinical endpoints, PROs, and follow-up 4DCTs will enable us to perform a complete evaluation of the ability of the novel computational tools to predict post-surgical lung function and side effects.

[0136] Hypothesis and Specific Aims

[0137] This project can test the central hypothesis that the novel, image-based lung-function evaluation methods as described herein can improve the accuracy of post-surgical lung-function prediction from 75% with existing methods to > 85% using the developed techniques. The central hypothesis can be tested via 2 specific aims (SA). SA 1 can assess the accuracy of 4DCT- ventilation / perfusion in predicting global post-surgical lung function using PFTs, clinical end points, and PROs. SA 2 can develop novel image processing and ML methods to analyze and predict complex changes in local lung function based on pre-surgical 4DCT-ventilation / perfusion imaging. Each aim is designed to be standalone where the success of each aim is independent of the success of the other.

[0138] Evaluate the Accuracy of Novel Lung-function Imaging to Predict Global Surgical Outcomes

[0139] Compare global lung-function measures predicted from 4DCT-ventilation / perfusion to conventional, anatomic approximation: The %PPO PFT metric can be used to evaluate global prediction accuracy. %PPO PFT is the accepted clinical standard to evaluate pre-surgical lung functions. The %PPO PFT metric predicts post-surgical PFTs by scaling the pre-surgery PFT value by the amount of lung function that will be lost by resection, where the amount of lung function lost by resection is calculated using an anatomic approximation of the surgical volume (Equation 1). Both novel 4DCT-ventilation / perfusion and conventional anatomic methods can be used to calculate %PPO PFT metrics. The novel 4DCT-ventilation / perfusion calculation will scale the pre-surgery PFT by the amount of ventilation / perfusion in the surgical volume as in Equation 3 : where PFT0is the pre-surgical PFT metric, {Ventilation or Perfusion {removed is the ventilation / perfusion measured in the surgical volume and {Ventilation or Perfusion}total is the ventilation / perfusion in the total lung volume. The novel and conventional calculated %PPO PFT metrics can be compared against the true, postoperative PFTs. The comparison of functional imaging-based %PPO PFTs metrics to the postoperative PFTs can provide a direct, clinically relevant assessment of novel 4DCT-ventilation / perfusion evaluation of lung-function.

[0140] Evaluate accuracy of 4DCT-ventilation / perfusion in predicting post-surgical clinical toxicity: This study dataset includes clinical evaluations of post-surgical complications, liters of 02 the patient requires, and the arterial 02 saturation. The rate of surgical complications and the change in each clinical feature from pre- to post-surgery can be calculated. The change in clinical feature can be used as an endpoint and to evaluate whether 4DCT-ventilation / perfusion-based %PPO PFT metrics predict for change in the clinical feature more accurately than the anatomic calculation of %PPO PFT. The clinical toxicity endpoints can allow for the assessment of the accuracy of novel 4DCT-ventilation / perfusion in predicting post-surgical lung function using clinically relevant measures of lung toxicity.

[0141] Evaluate accuracy of 4DCT-ventilation / perfusion for predicting patient-reported outcomes: Patients filled out a Quality of Life (QOL) questionnaire prior to surgery and 3 months after surgery. The study used the Functional Assessment of Cancer Therapy-Lung (FACT-L) questionnaire, a self-report instrument that measures 5 components of quality of life (physical, social / family, emotional and functional wellbeing, and disease-related concerns). Evaluation of whether novel (4DCT-ventilation / perfusion-based) or conventional (anatomic) %PPO PFT can be done to better predict for clinically meaningful changes in total QOL score (where a change of > 2 points is defined as clinically meaningful). Patient-reported QOL metrics have been cited as ‘possibly the most important outcome measures’ for major thoracic clinical trials. The QOL thoracic questionnaire will enable us to evaluate the accuracy of novel 4DCT- ventilation / perfusion in predicting post-surgical lung function using patient-centered evaluations (i.e., PROs).

[0142] Significance and challenges of SA 1 : The results from SA 1 can provide simple, clear data on the accuracy of 4DCT-ventilation / perfusion for predicting post-surgery lung function. Prediction of post-surgical lung function using a variety of measures and evaluation of the prediction accuracy with true, post-surgical evaluations can be performed. The data from SA 1 can provide clear evidence-based results on whether 4DCT-ventilation / perfusion is superior in evaluating pre- surgical lung function compared to conventional methods.

[0143] Develop Novel Methods for Local Functional Evaluation Post Lung-cancer Resection Characterize changes in spatial lung function following surgical resection for lung cancer: The expected pattern of functional change based on the %PPO PFT formula is that the spatial lung function post-surgery should be identical to the spatial lung function pre-surgery except for the lost function due to the resected tissue. However, studies have revealed that lung function may not redistribute equally after surgery. The 4DCT imaging collected before and after resection enables us to evaluate anatomic and functional imaging changes associated with surgery. FIG. 12 shows an example of 4DCT-ventilation / perfusion imaging collected before and after surgical resection of the right-lower lobe for a single patient. FIG. 13 is a schematic diagram of a Specific Aim (SA) evaluations and comparisons, according to some embodiments. FIG. 13 are illustrative representations depicting 4-dimensional computed tomography for pre-surgery and post-surgery, according to some embodiments. As shown in FIG. 13, example 4DCT-ventilation (left) and - perfusion (right) imaging are overlaid on a CT. Images can be collected from a single patient before (top) and 3 months after (bottom) surgical resection of the right-lower lobe. Bright colors denoted in region 1304 can show functional lung while blues and dark tones as denoted in region 1308 can display regions of ventilation / perfusion defect. The images reveal that, for this patient, ventilation largely followed the expected pattern (i.e., no changes aside from the removed tissue). However, perfusion redistributed in a non-intuitive way after surgery. Evaluation of salient features of anatomic and functional images before and after surgery can be performed, as well as features of image dynamics. The anatomic image features can include total lung volume and lung segmental volume, among others. The functional image features can include the coefficient of variation and percent ventilation in each lobe. The image dynamics can be obtained using validated the deformable image registration (DIR) algorithm to map pre-surgical anatomy / function to post-surgical anatomy / function. Image-dynamic features can include the deformation vector field and the Jacobian determinant. The anatomic and functional imaging comparison can enable the characterization of how anatomy and function redistribute after resectional surgery.

[0144] Statistical Analysis

[0145] The primary endpoint can be the comparison of %PPO PFTs to the true postoperative PFTs (SA 1.1). Predictive accuracy with the Concordance Correlation Coefficient (CCC) can be evaluated. Sex as a biological variable can also be evaluated by evaluating CCC for males and females independently in addition to in the mixed-sex cohort. Similarly, the CCC metrics can be used for clinical (SA 1.2), PRO (SA 1.3), and ML-predicted image (SA 2.3) evaluations. Exploratoary analysis of the ability of the novel methods to alter clinical decision making can be done using receiver operating characteristic methods.

[0146] EQUIVALENTS Although preferred embodiments of the invention have been described using specific terms, such description is for illustrative purposes only, and it is to be understood that changes and variations may be made without departing from the spirit or scope of the following claims.

[0147] INCORPORATION BY REFERENCE

[0148] The entire contents of all patents, published patent applications, and other references cited herein are hereby expressly incorporated herein in their entireties by reference.

Claims

CLAIMS1. A non-transitory, processor-readable medium storing instructions that when executed by a processor, cause the processor to: receive user data that includes a plurality of images of a body and organ; generate, via computed tomography imaging, a first image map and a second image map based on the user data; analyze, via spatial normalization of each image from the plurality of images of the user data, the first image map and the second image map to determine strength of correlation of a tissue point from a plurality of tissue points of each image; execute a mechanistic or trained machine-learning model to generate a functional image map using the first image map and the second image map as inputs, the functional image map includes a contour of high functional regions of the organ based on a predetermined threshold for function; and combined with radiation dose predict a change in a patient reported outcome for a user associated with the user data based on the dose distribution image map.

2. The non-transitory, processor-readable medium of claim 1, wherein the instructions to cause the processor to generate the first image map comprise instructions to further cause the processor to: segment the plurality of images of the organ to identify the organ while excluding pulmonary vasculatures and airways of the body during an inhale phase and an exhale phase of the user data; map corresponding tissue points on the organ in each image from the plurality of images of the user data in phase during the inhale phase and the exhale phase; and calculate function in each tissue point from the plurality of tissue points in each image from the plurality of images to generate the first image map and the second image map.

3. The non-transitory, processor-readable medium of claim 2, wherein the instructions to cause the processor to map corresponding tissue points include instructions to further cause the processor to capture changes in ventilation and perfusion throughout a breathing cycle.

4. The non-transitory, processor-readable medium of claim 1, wherein the instructions to cause the processor to analyze the first image map and the second image map include instructions to further cause the processor to execute a voxel-based lung-function analysis.

5. The non-transitory, processor-readable medium of claim 4, wherein the voxel -based lungfunction analysis can be executed via machine-learning.

6. The non-transitory, processor-readable medium of claim 1, wherein the first image map is a 4D computed tomography ventilation image map.

7. The non-transitory, processor-readable medium of claim 1, wherein the second image map is a 4D computed tomography perfusion image map.

8. The non-transitory, processor-readable medium of claim 1, wherein the machine-learning model is at least one of a supervised machine-learning model and an unsupervised machinelearning model.

9. The non-transitory, processor-readable medium of claim 1, wherein the trained machinelearning model is trained using a training set that includes a pre-treatment image correlated to a post-treatment dose distribution image map.

10. The non-transitory, processor-readable medium of claim 1, wherein the predetermined threshold for ventilation is 15% or greater ventilation.

11. An apparatus, comprising: a processor; and a memory operatively coupled to the processor, the memory storing instructions to cause the processor to: receive user data that includes a plurality of images of a body and organ;generate, via computed tomography imaging, a first image map and a second image map based on the user data; analyze, via spatial normalization of each image from the plurality of images of the user data, the first image map and the second image map to determine strength of correlation of a tissue point from a plurality of tissue points of each image; execute a mechanistic or trained machine-learning model to generate a functional image map using the first image map and the second image map as inputs, the functional image map includes a contour of high functioning regions of the organ based on a predetermined threshold for ventilation; and predict a change in a patient reported outcome for a user associated with the user data based on the dose distribution and functional image map.

12. The apparatus of claim 11, wherein the instructions to cause the processor to generate the first image map comprise instructions to further cause the processor to: segment the plurality of images of the organ to identify the organ while excluding pulmonary vasculatures and airways of the body during an inhale phase and an exhale phase of the user data; map corresponding tissue points on the organ in each image from the plurality of images of the user data in phase during the inhale phase and the exhale phase; and calculate ventilation in each tissue point from the plurality of tissue points in each image from the plurality of images to generate the first image map and the second image map.

13. The apparatus of claim 12, wherein the instructions to cause the processor to map corresponding tissue points include instructions to further cause the processor to capture changes in ventilation and perfusion throughout a breathing cycle.

14. The apparatus of claim 11, wherein the instructions to cause the processor to analyze the first image map and the second image map include instructions to further cause the processor to execute a voxel-based lung-function analysis.

15. The apparatus of claim 11 , wherein the first image map is a 4D computed tomography ventilation image map.

16. The apparatus of claim 11, wherein the second image map is a 4D computed tomography perfusion image map.

17. The apparatus of claim 11, wherein the machine-learning model is at least one of a supervised machine-learning model and an unsupervised machine-learning model.

18. The apparatus of claim 11, wherein the trained machine-learning model is trained using a training set that includes a pre-treatment image correlated to a post-treatment dose distribution image map.

19. The apparatus of claim 1, wherein the predetermined threshold for ventilation is 15% or greater ventilation.

20. A method comprising: combining dose and functional imaging; providing the combined dose and functional imaging information to a computer system including an artificial intelligence network, the artificial intelligence network having been trained using a plurality of patient information including dose and functional imaging from previous treatments and associated patient reported outcomes; and predicting a patient reported outcome for a target patient based on the computer system.

Images