Predicting outcomes in patients with endoluminal valves
The CVQS system uses machine learning to assess collateral ventilation, enhancing the effectiveness of lumen valve placement by identifying suitable candidates and predicting patient outcomes, addressing the ineffectiveness of lumen valves in high collateral ventilation scenarios.
Patent Information
- Application Number
- JP2024523749
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-10-20
- Filing Date
- 2022-10-19
- Publication Date
- 2024-11-06
- Estimated Expiration
- 2042-10-19
AI Technical Summary
Lumen valve placement may be ineffective for lung volume reduction in patients with chronic obstructive pulmonary disease (COPD) due to high collateral ventilation, as air can bypass the valves through collateral channels, reducing treatment efficacy.
A collateral ventilation quantification system (CVQS) using machine learning models to assess collateral ventilation, incorporating pressure and airflow sensors, and a CVQS device to determine the suitability of lumen valve placement by analyzing airflow and pressure data, providing indications of collateral ventilation presence and degree, and predicting patient outcomes.
The CVQS system accurately determines the presence and extent of collateral ventilation, enabling clinicians to identify suitable candidates for lumen valve placement, thereby improving treatment efficacy by ensuring effective lung volume reduction.
Smart Images

Figure 2024540959000001_ABST
Abstract
Description
[Technical field]
[0001] Claiming priority This application claims the benefit of priority to U.S. Provisional Application No. 63 / 262,776, entitled “ENDOLUMINAL VALVE PLACEMENT PATIENT OUTCOME PREDICTION,” filed on October 20, 2021, which is incorporated by reference in its entirety herein. [Background technology]
[0002] Luminal valves may be placed inside airways leading to diseased portions of the lungs to redirect breathed air from the diseased areas to healthier portions of the lungs. These lumenal valves are one-way valves that allow air and fluids (e.g., mucus) to escape from diseased portions of the lungs while preventing breathed air from entering the diseased portions of the lungs. Because diseased portions of the lungs (e.g., lung areas with significant emphysema) tend to increase in volume, preventing other portions from expanding properly, lumenal valve placement is an effective treatment to reduce the volume occupied by diseased portions of the lungs (which do not contribute to O2-CO2 gas exchange). Reducing the volume of the diseased portions provides the healthy portions of the lungs with more space to fully expand during the respiratory cycle, which allows significantly more gas exchange. Unfortunately, some diseased portions of the lungs may receive airflow from collateral ventilation, in which gas passes from one lung unit to an adjacent lung unit through collateral channels, such as alveolar stomas and / or direct airway anastomoses. Although luminal valve placement can be an effective treatment when some degree of collateral ventilation is present, when the degree of collateral ventilation is relatively high, luminal valve placement can be an ineffective treatment for lung volume reduction. Summary of the Invention [Means for solving the problem]
[0003] In the drawings, which are not necessarily drawn to scale, like numerals in different drawings may represent like components. Like numerals with different letter suffixes may represent different instances of like components. The drawings illustrate generally, by way of example, but not by way of limitation, various embodiments discussed in the present specification. [Brief description of the drawings]
[0004] [Figure 1] FIG. 1 illustrates an exemplary collateral ventilation quantification system (CVQS), in accordance with at least one example of the present disclosure. [Diagram 2] FIG. 1 is a diagram illustrating training of a machine learning model, in accordance with at least one example of the present disclosure. [Diagram 3] FIG. 1 is a machine learning model inference diagram, in accordance with at least one example of the present disclosure. [Figure 4A] FIG. 13 is a flow and pressure diagram illustrating collateral ventilation, in accordance with at least one example of the present disclosure. [Figure 4B] FIG. 13 is a flow and volume diagram illustrating collateral ventilation, in accordance with at least one example of the present disclosure. [Figure 4C] FIG. 13 is a flow and pressure diagram illustrating the absence of collateral ventilation (absence of CV), in accordance with at least one example of the present disclosure. [Figure 4D] FIG. 13 is a flow and volume diagram illustrating the absence of collateral ventilation (absence of CV), in accordance with at least one example of the present disclosure. [Diagram 5] 13 is a flowchart illustrating a technique for training a model to determine whether a patient is a candidate for receiving an endoluminal valve based on collateral ventilation data, in accordance with at least one example of the present disclosure. [Figure 6] 1 is a block diagram of an example machine in which any one or more of the techniques discussed herein may be performed, in accordance with at least one example of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0005] As mentioned above, some diseased portions of the lungs may receive airflow from collateral ventilation and / or collateral channels. Depending on the extent of it, this collateral ventilation (CV) may render luminal valve placement an ineffective treatment for lung volume reduction. This is because luminal valves may function properly and prevent air from entering the diseased portion of the lung through normal airways (e.g., bronchi, etc.), but collateral ventilation may exist to the extent that air freely enters the diseased portion without passing through the normal airways in which the valve is placed. Therefore, before treating a patient with chronic obstructive pulmonary disease (COPD) with luminal valve placement, the targeted lung region is usually evaluated to ensure that the COPD patient is not receiving airflow via collateral ventilation to the extent that the COPD patient is likely to not respond adequately to luminal valve placement.
[0006] A collateral ventilation quantification system (CVQS) may be used to assess the target lung region for CV. The CVQS may output sensor data (e.g., pressure and / or airflow) on a graphical user interface (GUI) or the like. As a specific, non-limiting example, measurements of pressure within the obstructed lobe and airflow into (or out of) the obstructed lobe may be sampled periodically or continuously. The output may include graphs and / or data points over time. A clinician (e.g., a doctor, surgeon, specialist, etc.) may evaluate the sensor data to determine whether CV is present in the target lung region and / or the extent of CV. However, in some instances, the clinician may not be able to determine whether CV is present, may take a long time to assess whether CV is present, and / or may not be able to accurately determine the extent of CV present with a high degree of confidence. In some instances, even if CV is present to some extent, the patient may still benefit from placement of an intraluminal valve, depending on various factors such as the patient's age, weight, body mass index (BMI), medical history, to name a few. However, the likelihood of a successful outcome with endoluminal valve placement cannot be determined by clinician assessment alone.
[0007] The systems and techniques described herein provide a model (e.g., a classifier, a trained model, etc., such as using machine learning techniques also known as artificial intelligence) to provide information related to whether a patient has CV. The model may output an indication that CV is present (CV+) or not present (CV-) in the patient's target lung region. In some examples, the model may output a probability and / or confidence level of whether CV is present, the degree of CV (e.g., an estimate of CV flow rate, such as low, medium, high, flow results over time, etc.), the degree of collateral resistance between the obstructed lung portion and adjacent lung portions, etc.
[0008] In one example, the model may output an indication of whether a patient is a suitable candidate for an intraluminal valve (instead of or in addition to whether CV is present). In this example, the indication may include a determination of CV, but may not be entirely dependent on whether CV is present. For example, a patient may benefit from an intraluminal valve in some examples despite having some CV.
[0009] 1 illustrates an exemplary collateral ventilation quantification system (CVQS) 100 according to at least one example of the present disclosure. Although the CVQS 100 is illustrated as a positive pressure system, other examples may not include positive pressure. For example, while the illustrated CVQS 100 includes airflow to the lung segment, other examples may include a system that allows airflow from the lung segment while preventing airflow to the lung segment. Generally speaking, in some examples, airflow into and out of the lung segment is restricted to determine whether collateral ventilation is present in the lung segment.
[0010] The CVQS 100 includes a CVQS device 102 and a CVQS tubing kit 104. The CVQS device 102 includes a flow meter 106, a pressure gauge 108, a display device 110 (e.g., including a graphical user interface), and a constant pressure air supply 112 (e.g., continuous positive airway pressure CPAP). In some examples, the CVQS device 102 may not include one or more of these components. For example, the constant pressure air supply 112 may not be used (and optionally, in this example, the flow meter 106 and the pressure gauge 108 may be omitted). In one example, the display device 110 may be located remotely from the CVQS device 102 (e.g., communicatively coupled via a wired or wireless communication architecture, such as Ethernet, Wi-Fi, Bluetooth, etc.).
[0011] The CVQS tubing kit 104 may include various tubes and / or filters, such as tubes used to add air to the lung segment and / or tubes to remove air from the lung segment. The CVQS tubing kit 104 includes a check valve 114 that may restrict air from entering or leaving the lung segment. In the example shown in the CVQS 100, the check valve 114 blocks airflow from the lung segment but allows air to enter the lung segment (e.g., via the constant pressure air supply 112). In other examples, the check valve may prevent airflow to the lung segment and allow air to exit the lung segment.
[0012] The CVQS 100 shown in FIG. 1 provides positive pressure to the occluded portion of the lung and measures pressure over time (using pressure gauge 108) and / or flow over time (using flow meter 106). Pressure and / or flow over time may be used to assess whether the occluded lung portion is ventilating air via collateral channels. The target lung portion may be occluded using an occlusion balloon (e.g., Olympus balloon catheter B7-2C), lumen valve, and / or other suitable occlusion device 116. In one example, the balloon catheter temporarily isolates a portion or portions of the lung by inflating in the airway. While inflated in the airway, the CVQS device 102 may provide airflow through the lumen of the catheter, for example, at a constant pressure (e.g., at 10 cmH2O). Airflow through the catheter lumen may be monitored by the flow meter 106 and / or pressure gauge 108 of the CVQS device 102, and data (e.g., flow, pressure, and / or total volume) may be output. The output may be displayed on a display device 110 (e.g., a portable computer, a mobile device, etc.). In some examples, the output is stored without being displayed. The output may be used to predict the patient's condition (e.g., whether or not they have collateral ventilation, and / or whether they are a good or bad candidate for an intraluminal valve), such as via a machine learning trained model.
[0013] A tubing kit 104 may connect the balloon catheter to the CVQS device 102. The tubing kit 104 may include a filter and a check valve to only allow airflow (e.g., supplied by the constant pressure air supply 112) to the target lobe, for example, for 1-10 minutes (e.g., 3-5 minutes), 5-20 minutes (e.g., about 10 minutes), etc. The flow through the balloon catheter may be affected by the amount of collateral ventilation in the tissue. For example, air may continue to flow into the obstructed lobe until the end-expiratory pressure equals the CVQS source pressure. If collateral ventilation is present, air continues to flow into the lobe as it escapes through the collateral channels (e.g., end pressure may not be reached, or may be reached and flow may continue). If air is able to escape from the obstructed lobe to an adjacent lobe via the collateral ventilation channels, pressure in the lobe may not equalize and / or air may continue to flow into the obstructed lobe, such as through the obstructed catheter lumen, and / or out of the obstructed lobe via the collateral ventilation channels.
[0014] In some emphysema or other lung patients, the alveoli of the lungs may swell and not allow gas exchange. As the alveoli become larger, they compress the other lobes, reducing the amount of air exchanged. A lumen valve may be used to allow air to leave the lung portion but not enter, thereby reducing the lung volume of the diseased portion of the lung. This type of valve may alleviate the problems associated with some of the symptoms of these patients. However, if collateral ventilation is present, the valve may fail or become less effective if air leaks between the lobes through the collateral channels. Collateral ventilation may occur when air moves between the lobes and / or between portions of the lobes. Collateral ventilation may occur in emphysema patients. In some patients, the lung fissures break open and create a breakthrough between the lobes, resulting in collateral ventilation. In some instances, a fissure integrity score may be used to assess whether a patient is a suitable candidate for a lumen valve. For example, if a patient has a fissure integrity score of greater than about 90%, the lobes may be assessed as being sufficiently intact to allow placement of a valve within the patient's body. If the score is below 90%, there may be more uncertainty as to whether the endoluminal valve will be effective. The CVQS100 may be used in conjunction with a machine learning trained model to evaluate a patient to determine whether they would benefit from an endoluminal valve.
[0015] 2 is a diagram 200 of training a machine learning model, in accordance with at least one example of the present disclosure. Diagram 200 illustrates components and inputs for training a model 202 using machine learning.
[0016] Machine learning (ML) is an application that provides computer systems with the ability to perform tasks without being explicitly programmed by making inferences based on patterns found in data analysis. Machine learning explores the study and construction of algorithms, also referred to herein as tools, that can learn from existing data and make predictions about new data. Examples are presented for several machine learning tools, but the principles presented herein may be applied to other machine learning tools.
[0017] The machine learning algorithm uses data (e.g., action and / or interaction primitives, goal vectors, rewards, etc.) to find correlations between identified features that affect outcomes. A feature is an individually measurable characteristic of the phenomenon being observed. Exemplary features of the model 202 may include diagnostic data (e.g., from a physician), reported patient outcome data, with or without a lumen valve, and / or other labels regarding the patient's condition and / or status. Features, which may include and / or be referred to as label data, may be compared to input data, such as pressure data, flow data, etc.
[0018] The concept of features is related to the concept of explanatory variables used in statistical techniques such as linear regression. Selecting informative, discriminatory, and independent features is important for the effective operation of ML in pattern recognition, classification, and regression. Features can be of various types, such as numerical features, strings, and graphs.
[0019] During training, the ML algorithm analyzes input data based on the identified features and, optionally, configuration parameters defined for the training (e.g., patient data such as environmental data, condition data, demographics and / or comorbidities, etc.). The result of the training is a model 202, which can receive inputs to generate complex tasks.
[0020] In one example, the input data may be labeled (e.g., for use as features in the training phase). Labeling may include identifying a patient's condition and / or status after treatment and / or after no treatment. For example, the patient's condition and / or status may be labeled as including or not including endoluminal valve intervention. The patient's condition and / or status may include objective outcomes (e.g., whether CV was present or not, the patient's breathing improved based on objective testing, etc.) and / or subjective outcomes (e.g., the patient perceives an improvement in breathing and / or quality of life, the clinician assesses whether CV was present or not using visual determination, etc.). Outcomes may be identified over time, such as 3 months and / or 6 months after intervention (or without intervention). The time labels may be weighted and / or used to generate different versions of the model 202. An example of an objective outcome includes a crack integrity score. The score may be weighted, such as 90 or above indicating no CV and 80 or below indicating CV. In some examples, the labels may include a weighting for the degree of CV. In these examples, the weighting may be based on low or high flow rates to improve the model 202. Some of the results described herein may be used to update the model 202 after initial training.
[0021] Input training data for the model 202 may include pressure and / or flow data, as discussed above. Other data used for input data may include CT scans (e.g., high resolution) with corresponding crack integrity scores, disease status, patient age, confounding factors such as infection, etc.
[0022] Neural networks, also called artificial neural networks, are computing systems based on the consideration of biological neural networks in animal brains. Such systems gradually improve their performance to perform tasks, called learning, usually without task-specific programming. For example, in image recognition, a neural network can be trained to identify images containing objects by analyzing example images tagged with the names of the objects, and once it has learned the objects and names, it can use the analysis results to identify and / or classify the objects in untagged images. For example, in FIG. 2, model 202 can be trained to identify whether a patient has CV and / or classify a patient (e.g., as a candidate for an intraluminal valve or not, and / or along with the probability and / or confidence of success of an intraluminal valve procedure) based on input data (e.g., pressure and / or flow rate).
[0023] Neural networks are based on a collection of connected units called neurons, and each connection between neurons, called a synapse, can send a unidirectional signal with different activation strength depending on the strength of the connection. A receiving neuron can activate and propagate a signal to downstream neurons connected to it, and this is usually done based on whether the combined received signal, which could be a signal from many sending neurons, is strong enough, where the strength is a parameter.
[0024] A deep neural network (DNN) is a stacked neural network made up of multiple layers. A layer is made up of nodes, which are where computations take place, loosely patterned on the neurons of the human brain, and which fire when they encounter enough stimuli. A node combines an input from the data with a set of coefficients and / or weights to amplify or attenuate that input, assigning importance to the input for the task the algorithm is trying to learn. The products of these inputs and weights are summed, and the sum is passed to the so-called activation function of the node to determine whether and to what extent the signal should progress further in the network and affect the final outcome. DNNs use a cascade of many layers of nonlinear processing units for feature extraction and transformation. Each subsequent layer uses the output from the previous layer as input. Higher-level features are derived from lower-level features to form a hierarchical representation. The layers following the input layer can be convolutional layers that produce a feature map that is the result of filtering the input and is used by the next convolutional layer.
[0025] The DNN may be a specific type of DNN, such as a convolutional neural network (CNN), a recurrent neural network (RNN), a long short-term memory (LSTM), etc. In some examples, other artificial neural networks may be used. In some examples, a classifier may be used instead of a neural network. A classifier may not include a hidden layer, but may classify a particular input as corresponding to a particular output. For example, for a set of pressure and / or flow data, a CV or non-CV identification may be generated by a classifier.
[0026] Input data for training model 202 may include data captured from a CVQS (e.g., CVQS 100 of FIG. 1 ) along with labeled data from a physician and / or patient. Model 202 may be used in an inference stage (described in further detail below with respect to FIG. 3 ) to determine the presence (or lack thereof) of collateral ventilation and / or to indicate whether a patient is a candidate for an endoluminal valve.
[0027] As shown in FIG. 2, the training data may include signal training data comprised of measurement signals representing quantifiable measurements obtained by the flowmeter and / or pressure sensor of the CVQS. For example, the training data may include measurements of airflow to and / or from a lobe that is occluded by the CVQS and is being considered for treatment by lumen valve placement in one or more airways in fluid communication with the occluded lobe. It will be understood that because the lobe is occluded, airflow into and out of the occluded lobe via the occluded air passage will pass completely through the lumen of the CVQS bypassing the occluding device (e.g., assuming the occluding device forms a perfect seal, which in practice does not always occur in real-life environments). The signal training data may include data provided by a CPAP machine (e.g., the constant pressure air supply 112 of FIG. 1) that is actively ventilating the patient. In some examples, the training data may include annotation training data provided by a physician and / or patient (e.g., as labeling data). For example, a physician can annotate a patient-specific CVQS dataset as CV present or absent, and / or as representing an assessment of CV and / or collateral resistance (Rcoll). This annotated training data can be used to train the model 202. Additionally or alternatively, the patient-specific CVQS dataset can be annotated with subjective patient outcome feedback following valve placement or not. For example, an individual patient can be evaluated via CVQS operation to generate a patient-specific CVQS dataset. Following this evaluation, one or more endoluminal valves can then be placed in the patient, and following this valve placement procedure (e.g., one week after valve placement, one month after valve placement, etc.), the patient can provide subjective feedback on perceived improvement or lack of improvement.Patient outcome feedback may include a sliding scale value of the perceived improvement in longevity resulting from valve placement (e.g., on a scale of 1 to 10 how previously reported shortness of breath improved following the valve placement procedure).
[0028] Based on the signal training data and / or the annotation training data, the model 202 may generate output weights corresponding to individual processing nodes spread across an input layer, an output layer, and one or more hidden layers. The model 202 and the trained weights may later be used to predict a CV+ or CV- indication, a degree of CV indication, and / or the patient's suitability for a predetermined type of treatment (e.g., endoluminal valve placement) based on new input from the patient under consideration.
[0029] 3 illustrates a machine learning model inference diagram 300 according to at least one example of the present disclosure. In the inference diagram 300, a model 302 (e.g., model 202 after training and / or upon update) may be used to output a prediction such as whether a patient has CV, whether a treatment is recommended (e.g., an intraluminal valve), etc. A confidence level and / or weighting may be output as a prediction or in addition to other predictions discussed above. The machine learning model inference diagram 300 may represent an exemplary computer-based clinical decision support system (CDSS) configured to assist in predicting a patient-specific outcome resulting from placement of one or more intraluminal valves in one or more bronchial airways in fluid communication with diseased portions of the lungs of a COPD patient.
[0030] As shown in FIG. 3, the model 302 may receive as inputs signals from a CVQS (e.g., CVQS 100 of FIG. 1), such as pressure data, flow data, and / or other data, such as data provided by a CPAP machine (e.g., constant pressure air supply 112 of FIG. 1) and / or patient data. The model 302 may generate outputs (e.g., inferences) including a patient suitability assessment (e.g., CV present or absent), a predicted patient outcome (e.g., an indication of the likely patient outcome if valve placement were performed based on the currently observed input signals), a confidence level (e.g., 95% confidence in a patient suitability assessment of "CV absent," 95% confidence that placement of the valve will reduce shortness of breath, etc.), an estimated amount of CV, etc. The model 302 may have a run time that occurs while the patient is undergoing and / or immediately following completion of the CVQS procedure. Model 302 may provide a physician with a rapid on-site assessment of whether a current patient is collaterally ventilated and / or whether a current patient could benefit from placement of a valve in an obstructed lung lobe.
[0031] FIG. 4A illustrates a flow and pressure diagram showing an example of collateral ventilation, according to at least one example of the present disclosure. FIG. 4A includes an arrow indicating whether balloon occlusion occurred (e.g., at 20 seconds). Indications of CV are not easily visible to the untrained eye, and in some examples, may be undetectable even to a trained physician. In one example, the output graph shown in FIG. 4A may be used as an input to a machine learning trained model to determine whether the corresponding patient has CV. The model may use a classifier and / or a neural network to determine whether CV is present from the graph and / or the underlying data.
[0032] 4B shows a flow and volume diagram illustrating an example of collateral ventilation, according to at least one example of the present disclosure. The vertical graph portion represents flow data, and the increasing line represents total flow.
[0033] FIG. 4C shows a flow and pressure diagram illustrating an example of a lack of collateral ventilation (absence of CV) according to at least one example of the present disclosure. FIG. 4D shows a flow and volume diagram illustrating an example of a lack of collateral ventilation (absence of CV) according to at least one example of the present disclosure. In the absence of collateral ventilation, airflow weakens and / or stops when the pressure in the obstructed lobe reaches the original pressure. If the ventilator is paused while the evaluation is being performed, air will continue to flow into the lobe until it is pressurized to the CPAP pressure. The graphs shown in FIG. 4C-FIG. 4D represent the (CV negative, absence of CV) condition.
[0034] 5 shows a flowchart illustrating a technique 500 for training a model to determine whether a patient is a candidate for an endoluminal valve based on collateral ventilation data, according to at least one example of the present disclosure. Technique 500 may be performed by a processor by executing instructions stored in a memory.
[0035] Technique 500 includes an operation 502 of receiving data, e.g., captured by a sensor, based on pressure and / or airflow in a target portion of a patient's lung that is obstructed such that it cannot receive air through the lung's respiratory airway. The obstructed respiratory airway may be obstructed by a balloon, blocking the outflow airway. In this example, the received data may be pressure data, e.g., based on a positive pressure applied to the inflow airway. The applied positive pressure may include a constant pressure, e.g., a constant pressure of 10 cmH2O. In some examples, the obstructed respiratory airway is obstructed by a valve that blocks the inflow airway while allowing outflow air, and the received data is outflow air data. In one example, operation 502 includes periodically obtaining measurement data of airflow and / or pressure in the target portion of the lung.
[0036] The technique 500 includes an act of labeling 504 the received data based on a corresponding patient respiratory outcome to generate training data and / or receive labeled data. In one example, the corresponding patient respiratory outcome includes a clinician's determination of whether the patient is collaterally ventilating the target portion of the lung based on the received data. In another example, the corresponding patient respiratory outcome includes objective metrics of the patient's breathing and / or a patient-reported respiratory assessment obtained after a procedure to insert an endoluminal valve within the patient. In some examples, a combination of corresponding patient respiratory outcomes may be used.
[0037] Technique 500 includes an operation 506 for training a machine learning model based at least in part on the training data to predict a patient's respiratory outcome via an indication of whether collateral ventilation is present in a target lung portion of a particular patient. Operation 506 may include using at least one of the volume data, the patient's medical images, the fissure integrity score, the patient's disease state, the patient's age, and / or the patient's comorbidities as additional input data. The indicia may include a binary indication of either collateral ventilation positive or collateral ventilation negative (e.g., via a user interface displaying text and / or images, via a light such as a green light for CV positive or a red light for CV negative). The indicia may include a likelihood that the patient is collateral ventilation positive, a confidence level, etc.
[0038] Technique 500 includes an operation 508 for outputting the machine learning model. Operation 508 can include deploying the machine learning model (e.g., making the machine learning model available via an API, the Internet, download, etc.), storing the machine learning model (e.g., for retrieval and / or updating for later use), sending the machine learning model to a destination (e.g., a database and / or server), etc.
[0039] The technique 500 may include an act to occlude a respiratory airway in a target portion of the lung (eg, using a valve, a balloon, etc.).
[0040] FIG. 6 is a block diagram of an example machine 600 on which any one or more of the techniques discussed herein may be performed, according to some embodiments. In alternative embodiments, the machine 600 may operate as a standalone device and / or may be connected (e.g., networked) to other machines. In a networked deployment, the machine 600 may operate in the capacity of a server machine, a client machine, or both, in a server-client network environment. In one example, the machine 600 may function as a peer machine in a peer-to-peer (P2P) (or other distributed) network environment. The machine 600 may be a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a network router, a switch or bridge, or any machine capable of executing instructions (sequential or otherwise) that specify actions that the machine should take. Additionally, although only a single machine is shown, the term "machine" is intended to include any collection of machines that individually or collectively execute a set (or sets) of instructions to perform any one or more of the methodologies described herein, such as cloud computing, software as a service (SaaS), other computer cluster configurations, etc.
[0041] The machine (e.g., computer system) 600 may include a hardware processor 602 (e.g., a central processing unit (CPU), a graphics processing unit (GPU), a hardware processor core, or any combination thereof), a main memory 604, and a static memory 606, some or all of which may communicate with each other via an interlink (e.g., a bus) 608. The machine 600 may further include a display unit 610, an alphanumeric input device 612 (e.g., a keyboard), and a user interface (UI) navigation device 614 (e.g., a mouse). In one example, the display unit 610, the input device 612, and the UI navigation device 614 may be touch screen displays. The machine 600 may further include a storage device (e.g., a drive unit) 616, a signal generating device 618 (e.g., a speaker), a network interface device 620, and one or more sensors 621, such as a global positioning system (GPS) sensor, a compass, an accelerometer, or other sensors. The machine 600 may include an output controller 628, such as a serial (e.g., Universal Serial Bus (USB)), parallel, or other wired or wireless (e.g., infrared (IR), near field communication (NFC), etc.) connection for communicating with and / or controlling one or more peripheral devices (e.g., printers, card readers, etc.).
[0042] Storage device 616 may include machine-readable medium 622 on which is stored one or more sets of data structures or instructions 624 (e.g., software) that embody or are utilized by any one or more of the techniques or functions described herein. The instructions 624 may also reside, completely or at least partially, within main memory 604, static memory 606, or within hardware processor 602 during execution by machine 600. In one example, one or any combination of hardware processor 602, main memory 604, static memory 606, or storage device 616 may constitute a machine-readable medium.
[0043] Although the machine-readable medium 622 is depicted as a single medium, the term "machine-readable medium" may include a single medium or multiple media (e.g., centralized or distributed databases and / or associated caches and servers) configured to store one or more instructions 624. The term "machine-readable medium" may include any medium that can store, encode, or carry instructions for execution by the machine 600, cause the machine 600 to perform any one or more of the techniques of this disclosure, or store, encode, or carry data structures used by or associated with such instructions. Non-limiting examples of machine-readable media may include solid-state memory, optical media, and magnetic media.
[0044] The instructions 624 may be further transmitted or received over a communications network 626 using a communications medium via the network interface device 620 utilizing any one of a number of transport protocols (e.g., Frame Relay, Internet Protocol (IP), Transmission Control Protocol (TCP), User Datagram Protocol (UDP), Hypertext Transfer Protocol (HTTP), etc.). Exemplary communications networks may include, among others, local area networks (LANs), wide area networks (WANs), packet data networks (e.g., the Internet), mobile telephone networks (e.g., cellular networks), Plain Old Telephone (POTS) networks, and wireless data networks (e.g., the Institute of Electrical and Electronics Engineers (IEEE) 802.11 family of standards known as Wi-Fi®, the IEEE 802.16 family of standards known as WiMax®), the IEEE 802.15.4 family of standards, peer-to-peer (P2P) networks, etc. In one example, the network interface device 620 may include one or more physical jacks (e.g., Ethernet, coaxial, or phone jacks) or one or more antennas for connecting to the communications network 626. In one example, the network interface device 620 may include multiple antennas for communicating wirelessly using at least one of single-input multiple-output (SIMO), multiple-input multiple-output (MIMO), or multiple-input single-output (MISO) techniques. The term "transmission medium" is intended to include any intangible medium capable of storing, encoding, or carrying instructions for execution by the machine 600, including digital or analog communications signals, or other intangible media that facilitate communication of such software.
[0045] Each of the following non-limiting examples may stand alone by itself or may be combined with one or more of the other examples in various permutations or combinations.
[0046] Example 1 is a collateral ventilation quantification system for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, the collateral ventilation quantification system comprising at least one sensor for capturing data based on at least one of pressure or airflow in a target portion of a patient's lung that is obstructed by a device such that it cannot receive air through the lung's respiratory airway; a processing circuit; and a memory containing instructions that, when executed by the processing circuit, cause the processing circuit to perform operations including labeling the received data based on a corresponding patient's respiratory outcome to generate training data; training a machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present in a particular patient's target lung portion; and storing the machine learning model.
[0047] Example 2 is a method for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, the method including receiving data captured by at least one sensor indicating at least one of pressure or airflow in a target portion of a patient's lung that is obstructed by a device such that it cannot receive air through the lung's respiratory airways, labeling the received data based on corresponding patient respiratory outcomes to generate training data, training a machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present in a particular patient's target lung portion, and outputting the machine learning model.
[0048] In Example 3, the subject matter of Example 2 includes the obstructed respiratory airway being occluded by a balloon to block the outflow airway, and the received data is pressure data based on positive pressure applied to the inflow airway.
[0049] In Example 4, the subject matter of Example 3 includes where the applied positive pressure includes a constant applied pressure.
[0050] In Example 5, the subject matter of Examples 2-4 includes training the machine learning model includes using at least one of lung segment volume data, a medical image of the patient, a fissure integrity score, a disease state of the patient, an age of the patient, or a comorbidity of the patient as additional input data.
[0051] In Example 6, the subject matter of Examples 2-5 includes where the corresponding patient respiratory outcome includes a clinician determination of whether the patient is experiencing collateral ventilation in a target portion of the lung based on the received data.
[0052] In Example 7, the subject matter of Examples 2-6 includes, wherein the corresponding patient respiratory outcome includes an objective outcome of the patient's breathing or a patient-reported respiratory assessment obtained following a procedure to insert an endoluminal valve within the patient's body.
[0053] In Example 8, the subject matter of Examples 2-7 includes, wherein an indication of whether collateral ventilation is present in a target lung portion of a particular patient is output from the model as a binary representation of whether collateral ventilation is present or not present.
[0054] In Example 9, the subject matter of Examples 2-8 includes outputting an indicia from the model that includes a probability that the patient is collaterally ventilating in the target segment.
[0055] In Example 10, the subject matter of Examples 2-9 includes using a device to occlude a respiratory airway in a target portion of the lung.
[0056] In Example 11, the subject matter of Examples 2-10 further comprises receiving the data including repeatedly or periodically obtaining measurement data of airflow or pressure in the target portion of the lung.
[0057] In Example 12, the subject matter of Examples 2-11 includes where the obstructed respiratory airway is obstructed by a valve that blocks the inflow airway while allowing outflow air, and where the received data is outflow air data.
[0058] Example 13 is a method for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, the method including receiving pressure data captured by at least one sensor indicative of pressure at a target portion of a patient's lung that is obstructed by a device such that it cannot receive air through the lung's respiratory airways, labeling the received pressure data based on corresponding patient respiratory outcomes to generate training data, training the machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present at a particular patient's target lung portion, and storing the machine learning model.
[0059] Example 14 is a method for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, the method including receiving airflow data captured by at least one sensor indicative of airflow exiting a target portion of a patient's lung that is obstructed by a device such that it cannot receive air through the lung's respiratory airways, labeling the received airflow data based on corresponding patient respiratory outcomes to generate training data, training the machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present in a particular patient's target lung portion, and storing the machine learning model.
[0060] Example 15 is a device for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, the device comprising a processing circuit and a memory including instructions that, when executed by the processing circuit, cause the processing circuit to perform operations including receiving data captured by at least one sensor indicating pressure or airflow at a target portion of a patient's lung that is obstructed by the device such that it cannot receive air through the lung's respiratory airway; labeling the received data based on corresponding patient respiratory outcomes to generate training data; training a machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present at a particular patient's target lung portion; and outputting the machine learning model.
[0061] Example 16 is at least one machine-readable medium including instructions that, when executed by a processing circuit, cause the processing circuit to perform operations including receiving data captured by at least one sensor indicating pressure or airflow in a target portion of a patient's lung that is obstructed by a device such that it cannot receive air through the lung's respiratory airway; labeling the received data based on a corresponding patient's respiratory outcome to generate training data; training a machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present in a particular patient's target lung portion; and outputting the machine learning model.
[0062] Example 17 is a method that includes receiving data captured by at least one sensor indicating at least one of pressure or airflow in a target portion of a patient's lung that is obstructed by a device such that it cannot receive air through the lung's respiratory airway; implementing a machine learning model trained at least in part based on training data including input previous patient sensor data and corresponding labeled previous patient respiratory outcomes to predict a patient respiratory outcome for the patient; and outputting an indication of at least one of whether collateral ventilation is present or whether the predicted patient respiratory outcome corresponds to placement of an endoluminal valve within the patient's body based on a prediction from the machine learning model.
[0063] In Example 18, the subject matter of Example 17 includes wherein the step of outputting the indicia includes the step of identifying that collateral ventilation is present and, in response, displaying a recommendation to treat the patient with an intraluminal valve.
[0064] In Example 19, the subject matter of Examples 17-18 includes wherein the step of outputting the indicia includes identifying that collateral ventilation is not present and, in response, displaying a recommendation not to treat the patient with an intraluminal valve.
[0065] In Example 20, the subject matter of Examples 17-19 includes where the obstructed respiratory airway is occluded by a balloon to block the outflow airway, and the received data is pressure data based on positive pressure applied to the inflow airway.
[0066] In Example 21, the subject matter of Example 20 includes where the applied positive pressure includes a constant applied pressure.
[0067] In Example 22, the subject matter of Examples 17-21 includes outputting the indicia including outputting a probability that the patient is collaterally ventilating in the target segment.
[0068] In Example 23, the subject matter of Examples 17-22 includes using a device to occlude a respiratory airway in a target portion of the lung.
[0069] In Example 24, the subject matter of Examples 17-23 further includes the step of receiving data including repeatedly or periodically obtaining measurement data of airflow or pressure in the target portion of the lung.
[0070] In Example 25, the subject matter of Examples 17-24 includes where the obstructed respiratory airway is obstructed by a valve that blocks the inflow airway while allowing outflow air, and where the received data is outflow air data.
[0071] Example 26 is at least one machine-readable medium including instructions that, when executed by a processing circuit, cause the processing circuit to perform operations to implement any of Examples 1-25.
[0072] Example 27 is an apparatus comprising means for implementing any of Examples 1 to 25.
[0073] Example 28 is a system for implementing any of Examples 1 to 25.
[0074] Example 29 is a method for implementing any of Examples 1 to 25.
[0075] The method examples described herein may be at least partially machine- or computer-implemented. Some examples may include computer-readable or machine-readable media encoded with instructions operable to configure an electronic device to perform the methods described in the above examples. Such method implementations may include code, such as microcode, assembly language code, high-level language code, and the like. Such code may include computer-readable instructions for performing various methods. The code may form part of a computer program product. Further, in one example, the code may be tangibly stored on one or more volatile, non-transitory, or non-volatile tangible computer-readable media during execution or at other times. Examples of these tangible computer-readable media may include, but are not limited to, hard disks, removable magnetic disks, removable optical disks (e.g., compact disks and digital video disks), magnetic cassettes, memory cards or sticks, random access memory (RAM), read-only memory (ROM), and the like. [Explanation of symbols]
[0076] 100 Collateral Ventilation Quantification System (CVQS) 102 CVQS Device 104 CVQS Tubing Kit 106 Flow meter 108 Pressure Gauge 110 Display Devices 112 Constant pressure air supply device 114 Check valve 116 Occlusion Devices 200 Machine Learning Model Training Diagrams 202 Model 300 Machine Learning Model Inference Diagram 302 Model 500 techniques 502 operation 600 machines 602 Hardware Processor 604 Main Memory 606 Static Memory 608 Interlink 610 Display Unit 612 Alphanumeric Input Device 614 User Interface (UI) Navigation Devices 616 Storage Devices 618 Signal Generating Device 620 Network Interface Device 621 Sensors 622 Machine-readable media 624 command 626 Communication Network 628 Output Controller
Claims
1. 1. A collateral ventilation quantification system for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, the collateral ventilation quantification system comprising: at least one sensor for capturing data based on at least one of pressure or airflow in a target portion of the patient's lung that is occluded by the device such that it cannot receive air through a respiratory airway of the lung; A processing circuit; When executed by the processing circuitry, the processing circuitry labeling the received data based on a corresponding patient's respiratory outcome to generate training data; training a machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present in a target lung portion of a particular patient; storing the machine learning model; a memory containing instructions for performing operations including A system comprising:
2. 1. A method for training a machine learning model for use in a computer-based clinical decision support system to assist in predicting patient outcomes of endoluminal valve placement, comprising: receiving data captured by at least one sensor indicative of at least one of pressure or airflow in a target portion of the patient's lung that is occluded by the device such that it cannot receive air through a respiratory airway of the lung; labelling the received data based on corresponding patient respiratory outcomes to generate training data; training a machine learning model based at least in part on the training data to predict one or more patient respiratory outcomes via an indication of whether collateral ventilation is present in a target lung portion of a particular patient; outputting the machine learning model; A method comprising:
3. 3. The method of claim 2, wherein the obstructed respiratory airway is occluded by a balloon to block an outflow airway, and the received data is pressure data based on positive pressure applied to an inflow airway.
4. The method of claim 3 , wherein the applied positive pressure comprises a constant applied pressure.
5. 3. The method of claim 2, wherein training the machine learning model includes using at least one of lung segment volume data, medical images of the patient, fissure integrity score, disease state of the patient, age of the patient, or comorbidities of the patient as additional input data.
6. The method of claim 2 , wherein the corresponding patient respiratory outcome comprises a clinician determination of whether the patient is experiencing collateral ventilation in the target portion of the lung based on the received data.
7. 3. The method of claim 2, wherein the corresponding patient respiratory outcome comprises an objective outcome of the patient's breathing or a patient-reported respiratory assessment obtained following a procedure to insert an endoluminal valve within the patient's body.
8. 3. The method of claim 2, wherein the indication of whether collateral ventilation is present in a target lung portion of a particular patient is output from the model as a binary representation of collateral ventilation being present or not being present.
9. The method of claim 2 , wherein the indicia output from the model includes a probability that the patient has collateral ventilation in the target portion.
10. The method of claim 2, further comprising using the device to occlude the respiratory airways of the target portion of the lung.
11. 11. A method according to any one of claims 2 to 10, wherein the step of receiving data comprises the step of repeatedly or periodically obtaining measurement data of the airflow or the pressure in the target portion of the lung.
12. 11. The method of any one of claims 5 to 10, wherein the obstructed respiratory airway is obstructed by a valve that blocks the inflow airway while allowing outflow air, and the received data is outflow air data.
13. receiving data captured by at least one sensor indicative of at least one of pressure or airflow in a target portion of the patient's lung that is occluded by the device such that it cannot receive air through a respiratory airway of the lung; implementing a machine learning model trained based at least in part on training data including the input prior patient sensor data and corresponding labeled prior patient respiratory outcomes to predict a patient respiratory outcome for the patient; outputting at least one indication of whether collateral ventilation is present or whether the predicted patient respiratory outcome corresponds to placement of an endoluminal valve within the patient based on the prediction from the machine learning model; A method comprising:
14. 14. The method of claim 13, wherein outputting the indicia comprises identifying that collateral ventilation is present and, in response, displaying a recommendation to treat the patient with the endoluminal valve.
15. 14. The method of claim 13, wherein outputting the indicia comprises identifying an absence of collateral ventilation and, in response, displaying a recommendation not to treat the patient with the endoluminal valve.
16. 14. The method of claim 13, wherein the obstructed respiratory airway is occluded by a balloon to block an outflow airway, and the received data is pressure data based on positive pressure applied to an inflow airway.
17. The method of claim 16 , wherein the applied positive pressure comprises a constant applied pressure.
18. The method of claim 13 , wherein outputting the indicia comprises outputting a probability that the patient is collaterally ventilating in the target segment.
19. The method of claim 13, further comprising using the device to occlude the respiratory airways of the target portion of the lung.
20. 20. A method according to any one of claims 13 to 19, wherein the step of receiving data comprises repeatedly or periodically obtaining measurement data of the airflow or the pressure in the target portion of the lung.
Citation Information
Patent Citations
Methods and systems for lung segment diagnosis
JP2009501568A
Direct lung sensor system, method, and apparatus
JP2011523363A
Collateral ventilation assessment system
JP2020189090A
