According to an embodiment, the ventilator system receives or otherwise determines one or more other input features utilized as input by the trained asynchrony detection algorithm. Other examples of input features received by the ventilator system and utilized by the trained asynchrony detection algorithm include, but are not limited to, one or more of an estimate of cardiogenic artifact within the breath, ventilator settings (e.g., threshold value, control mode, positive end expiratory pressure (PEEP) level, etc.), ventilator breathing circuit (e.g., single limb, dual limb, mask . . . ), excursion of patient effort (i.e., area under the ventilator threshold when the breath is triggered), slope of the patient effort in the pressure and/or flow channel, patient effort lowest/highest value in the pressure and/or flow channel (including around breath initiation), estimation of patient effort in the airway pressure channel, estimation of patient effort in the flow channel, estimate of intrinsic PEEP level (i.e., baseline level—PEEP), and estimate of patient respiratory mechanics (i.e., respiratory compliance and/or resistance), among other possible input features.

Once the ventilator system receives, retrieves, or otherwise determines the airway pressure waveform, the flow waveform, and/or one or more additional input features from the ventilator system, this information can be utilized immediately or can be stored in local or remote storage for subsequent use.

At step 130 of the method, the ventilator system calculates an airway pressure water accumulation index from the received airway pressure waveform for a patient breath. The ventilator system can calculate the airway pressure water accumulation index using a variety of different methods, and optionally calculates the airway pressure water accumulation index on a breath-by-breath basis. According to an embodiment, the ventilator system calculates the airway pressure water accumulation index on a breath-by-breath basis using the past one second of waveform data, although other timeframes and windows are possible.

Referring to FIG. 5, in one embodiment, is a method 500 for calculating an airway pressure water accumulation index from a received airway pressure waveform for a patient breath. At step 132 of method 500, the received airway pressure waveform is standardized so that there are zero mean and unit variance. According to an embodiment, standardizing the received airway pressure waveform comprises subtracting a calculated airway pressure waveform mean and dividing the received airway pressure waveform by a calculated standard deviation of the airway pressure waveform.

At step 134 of the method, the ventilator system determines an airway pressure waveform trend, comprising application of a median filter. The filter may utilize a variety of different sample sizes. According to an embodiment, the medial filter is applied with a kernel of size three samples (e.g., for waveforms sampled at 100 Hz).

At step 136 of the method, the ventilator system calculates an airway pressure waveform residue, comprising subtracting the determined waveform trend from the standardized airway pressure waveform for the breath.

At step 138 of the method, the ventilator system calculates a standard deviation of the computed airway pressure waveform residue. This comprises the final water accumulation index for the airway pressure waveform for the breath. Once it is determined, the airway pressure water accumulation index may be utilized immediately or may be stored in local or remote storage for subsequent use in the method.

At step 140 of the method, the ventilator system calculates a flow water accumulation index from the received flow waveform for a patient breath. The ventilator system can calculate the flow water accumulation index using a variety of different methods, and optionally calculates the flow water accumulation index on a breath-by-breath basis. According to an embodiment, the ventilator system calculates the flow water accumulation index on a breath-by-breath basis using the past one second of waveform data, although other timeframes and windows are possible.

Notably, according to an embodiment, steps 130 and 140 of the method are completed sequentially in any order, or the steps are completed in parallel.

Referring to FIG. 6, in one embodiment, is a method 600 for calculating a flow water accumulation index from a received flow waveform for a patient breath. At step 142 of method 600, the received flow waveform is standardized so that there are zero mean and unit variance. According to an embodiment, standardizing the received flow waveform comprises subtracting a calculated flow waveform mean and dividing the received flow waveform by a calculated standard deviation of the flow waveform. However, other methods for standardizing the received flow waveform are possible. For example, according to another embodiment, standardizing the received flow waveform comprises subtracting a calculated flow waveform median and dividing the received flow waveform by a calculated standard deviation of the flow waveform. Many other methods are possible.

At step 144 of the method, the ventilator system determines a flow waveform trend, comprising application of a median filter. The filter may utilize a variety of different sample sizes. According to an embodiment, the medial filter is applied with a kernel of size three samples (e.g., for waveforms sampled at 100 Hz).

At step 146 of the method, the ventilator system calculates a flow waveform residue, comprising subtracting the determined waveform trend from the standardized flow waveform for the breath.

At step 148 of the method, the ventilator system calculates a standard deviation of the computed flow waveform residue. This comprises the final water accumulation index for the flow waveform for the breath. Once it is determined, the flow water accumulation index may be utilized immediately or may be stored in local or remote storage for subsequent use in the method.

Referring to FIG. 7, in one embodiment, is another high-level overview 700 of a method 100 for automated detection of water accumulation in a ventilator circuit by a trained asynchrony detection algorithm of a ventilator system. At step 710, the input features for the trained asynchrony detection algorithm are determined or received. For example, the input features (feature 1, feature 2, . . . feature N) comprises at least the airway pressure water accumulation index and the flow water accumulation index, although the input features may comprise one or more additional input features.

Returning to FIG. 1, at step 150 of the method, the trained asynchrony detection algorithm of the ventilator system labels or otherwise identifies a single breath of the patient as synchronous or asynchronous. The trained asynchrony detection algorithm utilizes at least the airway pressure water accumulation index and the flow water accumulation index as input features, although the input features utilized by the trained asynchrony detection algorithm may comprise one or more additional input features. This label can be utilized immediately or can be stored in local or remote storage for subsequent use in the method.

According to an embodiment, the trained asynchrony detection algorithm of the ventilator system further labels or otherwise identifies the single breath of the patient as asynchronous and identifies that asynchrony as auto-triggered, double-triggered, or ineffectively triggered. Other types of asynchrony are possible.

According to an embodiment, the trained asynchrony detection algorithm can be any machine learning algorithm, classifier, or model sufficient to utilize the input features described or otherwise envisioned herein, and sufficient to label or otherwise identify a single breath of the patient as synchronous or asynchronous. According to an embodiment, the trained asynchrony detection algorithm is trained by the ventilator system. According to another embodiment, the trained asynchrony detection algorithm is trained by another system.

According to an embodiment, an asynchrony detection algorithm can be trained using breath-by-breath input feature such as airway pressure water accumulation index and the flow water accumulation index (along the other features) along with ground truth or other training data in order to be trained to predict or detect asynchronies, and label asynchronies as auto-triggered, double-triggered, or ineffectively triggered. Standard machine learning models can be used to train the asynchrony detection classifier. Training the asynchrony detection algorithm may comprise, for example, optimizing the kernel size, and/or exploring spectral approaches such as ratio of the first harmonic to the rest of the spectrum as a measure of high frequency content. According to one specific, non-limiting example, an asynchrony detection algorithm can be a Lasso model with 5-fold cross validation using a plurality of breaths from clinical and/or mechanical lab datasets.

Referring again to the high-level overview 700 of method 100 for automated detection of water accumulation in a ventilator circuit, at step 720 the trained asynchrony detection algorithm of the ventilator system labels or otherwise identifies the patient's breath as synchronous (e.g., “normal” or “no asynchrony”) or as asynchronous (e.g., auto-triggered (AT), double-triggered (DT), ineffectively triggered (IT), or another type of asychrony).

Returning to FIG. 1, at step 160 of the method, the trained asynchrony detection algorithm of the ventilator system ranks the plurality of input features in a ranked order, where the ranked order is based on an importance of each respective input feature to the determination of the single breath of the patient as synchronous or asynchronous by the trained asynchrony detection algorithm. Accordingly, the asynchrony detection algorithm of the ventilator system is further trained to determine the importance of each input feature to the output of the algorithm (i.e., the labeling of the single breath of the patient as synchronous or asynchronous), and to rank the input features according to that determined importance. The ranking therefore comprises at least a determination of a most important input feature.

As described or otherwise envisioned herein, the plurality of input features comprises at least the airway pressure water accumulation index and the flow water accumulation index as input features, although the plurality of input features may comprise one or more additional input features. Accordingly, the ranking may be a ranked order of at least the airway pressure water accumulation index and the flow water accumulation index.

Once the ranking of the plurality of input features and the determination of the most important input feature for the patient breath is determined, this output can be utilized immediately or can be stored in local or remote storage for subsequent use.

Referring again to the high-level overview 700 of method 100 for automated detection of water accumulation in a ventilator circuit, at step 730 the trained asynchrony detection algorithm of the ventilator system ranks the input features based on importance of the feature on the determination of synchronous or asynchronous for the patient breath. For example, in FIG. 7 the water accumulation index is the most important input feature for that patient breath, followed by input feature three and then by input feature two.

According to an embodiment, multiple iterations of steps 120 through 160 are completed, one iteration each for each of a plurality of a patient's breaths. These plurality of breaths may be a plurality of or all consecutive breaths in a predetermined window or timeframe, or may be sampled non-consecutive breaths within a predetermined window or timeframe. The predetermined window or timeframe may be a setting of the ventilator system, a preprogrammed default, a variable predetermined window or timeframe determined by a parameter of the patient, the ventilator, and/or one or more ventilator settings, among other possibilities.

At step 170 of the method, the trained asynchrony detection algorithm of the ventilator system utilizes data from a plurality of iterations of step 120 through 160, thus data for a plurality of a patient's breaths, to determine several different parameters. As one determined parameter, the system determines a most frequent breath labeling for the plurality of iterations, wherein the most frequent breath labeling is one of synchronous (N), auto-triggered (AT), double-triggered (DT), ineffectively triggered (IT), or another type of asychrony. According to one possible embodiment, determining a most frequent breath labeling for the plurality of iterations comprises determining, for the plurality of iterations, a percentage of each of the synchronous, auto-triggered, double-triggered, ineffectively triggered, or other type of asychrony breath labels relative to all labels.

For example, referring to step 740 in FIG. 7, determining a most frequent breath labeling for the plurality of iterations comprises determining the percentage of synchronous (N 5%), auto-triggered (AT 15%), double-triggered (DT 75%), or ineffectively triggered (IT 5%) breath labels relative to all labeled breaths (100%).

According to an embodiment, as another determined parameter, the system determines a most frequent most important input feature for the plurality of iterations. According to one possible embodiment, determining a most frequent most important input feature for the plurality of iterations comprises determining, for the plurality of iterations, a percentage of iterations each of the input features is most important to the determination by the trained asynchrony detection algorithm.

For example, referring to step 740 in FIG. 7, determining a most frequent most important input feature for the plurality of iterations comprises determining that the water accumulation index was the most frequent most important input feature (80% of the iterations), while leak was the second most frequent most important input feature (10% of the iterations), and one or more other input features were the most important input feature the remaining percentage of the iterations (10%).

According to an embodiment, as another determined parameter, the system determines a most frequent most important input feature for each of the synchronous, auto-triggered, double-triggered, and ineffectively triggered breath labels. According to one possible embodiment, determining a most frequent most important input feature for each of the synchronous, auto-triggered, double-triggered, and ineffectively triggered breath labels comprises determining a percentage of the most important input feature for each of the breath labels.

For example, referring to step 740 in FIG. 7, determining a most frequent most important input feature for each of the synchronous, auto-triggered, double-triggered, and ineffectively triggered breath labels comprises determining that for the double-triggered (DT) breath labels (which was 75% of the breath labels as determined by the trained asynchrony detection algorithm above), the most frequent most important input feature was water accumulation which was 100% of the time. For auto-triggered (AT) breath labels (which was 15% of the breath labels as determined by the trained asynchrony detection algorithm above), the most frequent most important input feature was leak which was 95% of the time.

According to an embodiment, once the most frequent breath labeling for the plurality of iterations, the most frequent most important input feature for the plurality of iterations, and the most frequent most important input feature for each of the synchronous, auto-triggered, double-triggered, and ineffectively triggered breath labels are determined, these parameters may be utilized immediately or may be stored in local or remote storage for subsequent use in the method.

Thus, according to an embodiment, information provided to the user is be divided into three levels: (i) a ranked list of labels with the corresponding percentage of occurrence (frequency) (e.g., 75% DT, 15% AT . . . ); (ii) a ranked list of features with the corresponding percentage of times they ranked highest in importance (e.g., water accumulation index ranked highest 80% of all breaths); and (iii) a ranking of features by importance and provided label by label to relate the root cause of asynchrony. For example, as shown in FIG. 7, DT occurred in 75% of the breaths, and in all of those breaths the highest-ranking feature was the water accumulation index. Thus, in this case DT was mainly caused by water accumulation. Notably, there can be cases where DT is caused by another major reason or there is no asynchrony but still there is water accumulation. TABLE 1 identifies several of these differences and how they affect the notifications.

Referring to TABLE 1 are several case examples. According to an embodiment, the trained asynchrony detection algorithm can capture root causes of asynchrony (i.e., if water accumulation is a root cause) but at the same time can protect a user from incorrect conclusions. For example, DT can be high and water accumulation could be present; however, it might not be the major reason for DT. For example, in case 2 even though there is no asynchrony water accumulation still exists because the water accumulation index is still ranked high in these features, or its value exceeds a certain threshold determined by the algorithm. In case 3, water accumulation exists but it is not causing the DT. While in Case 4, DT is specifically caused by the water collection.

TABLE 1

Case Examples

Water

Water

accumulation

accumulation

index ranking

index ranking
Notification on water

N
for N
DT
for DT
accumulation index

Case 1
↑
↓
↓
↓
No water accumulation in circuit.

No actions needed.

Case 2
↑
↑
↓
↓
Water accumulation detected -

remove water from circuit

Case 3
↑
↑
↑
↓
Water accumulation detected -

remove water from circuit. DT

detected but caused by Feature X

(not water accumulation)

Case 4
↓
↓
↑
↑
Water accumulation causing DT-

remove water from circuit to

reduce DT

N: normal breaths,

DT: double triggering breaths

Returning to method 100 in FIG. 1, at step 180 the trained asynchrony detection algorithm utilizes the determined parameters (e.g., the most frequent breath labeling for the plurality of iterations, the most frequent most important input feature for the plurality of iterations, and the most frequent most important input feature for each of the synchronous, auto-triggered, double-triggered, and ineffectively triggered breath labels) to determine one or more of: (i) water accumulation in the ventilator circuit and an action to remove the water accumulation from the ventilator circuit; and (ii) asynchrony in the ventilator circuit due to a cause other than water accumulation, and an action to correct the asynchrony.

According to an embodiment, therefore, the asynchrony detection algorithm is further trained to determine, from the provided parameters, whether there is water accumulation in the ventilator circuit. The asynchrony detection algorithm is further trained to determine, from the provided parameters, what action should be taken to remove the water accumulation from the ventilator circuit, or what action should be taken if there is asynchrony in the ventilator circuit due to a cause other than water accumulation.

Once the trained asynchrony detection algorithm utilizes the determined parameters to identify water accumulation and/or asynchrony as well as an action to take, this information may be utilized immediately or can be stored in local or remote storage for subsequent use in the method.

Referring again to the high-level overview 700 of method 100 for automated detection of water accumulation in a ventilator circuit, at step 750 the trained asynchrony detection algorithm of the ventilator system has determined that water accumulation is causing double-triggering (DT) and has determined an action to take to correct the situation (“remove water from circuit”).

Returning to method 100 in FIG. 1, at step 190 of the method, the ventilation system displays or otherwise provides one or both of: (i) the determination of water accumulation in the ventilator circuit and/or the determination of asynchrony due to a reason other than water accumulation; and (ii) the action determined to correct the water accumulation and/or asynchrony, using a user interface 240 of the ventilator system 200. The user interface may be any method or component sufficient to communicate the information to a user. According to an embodiment, the system may display a report on the user interface or display of the system, or may issue an alarm. The information may be communicated by wired and/or wireless communication to another device. For example, the system may communicate the report to a mobile phone, computer, laptop, wearable device, and/or any other device configured to allow display and/or other communication of the report.

Referring again to the high-level overview 700 of method 100 for automated detection of water accumulation in a ventilator circuit, at step 750 the trained asynchrony detection algorithm of the ventilator system has determined that water accumulation is causing double-triggering (DT) and has determined an action to take to correct the situation (“remove water from circuit”), and this information is communicated to the user via a user interface.

Referring again to FIG. 2 is a schematic representation of a ventilation system 200. System 200 may be any of the systems described or otherwise envisioned herein, and may comprise any of the components described or otherwise envisioned herein. It will be understood that FIG. 2 constitutes, in some respects, an abstraction and that the actual organization of the components of the system 200 may be different and more complex than illustrated.

According to an embodiment, system 200 comprises a processor 220 capable of executing instructions stored in memory 230 or storage 260 or otherwise processing data to, for example, perform one or more steps of the method. Processor 220 may be formed of one or multiple modules. Processor 220 may take any suitable form, including but not limited to a microprocessor, microcontroller, multiple microcontrollers, circuitry, field programmable gate array (FPGA), application-specific integrated circuit (ASIC), a single processor, or plural processors.

Memory 230 can take any suitable form, including a non-volatile memory and/or RAM. The memory 230 may include various memories such as, for example L1, L2, or L3 cache or system memory. As such, the memory 230 may include static random access memory (SRAM), dynamic RAM (DRAM), flash memory, read only memory (ROM), or other similar memory devices. The memory can store, among other things, an operating system. The RAM is used by the processor for the temporary storage of data. According to an embodiment, an operating system may contain code which, when executed by the processor, controls operation of one or more components of system 200. It will be apparent that, in embodiments where the processor implements one or more of the functions described herein in hardware, the software described as corresponding to such functionality in other embodiments may be omitted.

User interface 240 may include one or more devices for enabling communication with a user. The user interface can be any device or system that allows information to be conveyed and/or received, and may include a display, a mouse, and/or a keyboard for receiving user commands. In some embodiments, user interface 240 may include a command line interface or graphical user interface that may be presented to a remote terminal via communication interface 250. The user interface may be located with one or more other components of the system, or may be located remote from the system and in communication via a wired and/or wireless communications network.

Communication interface 250 may include one or more devices for enabling communication with other hardware devices. For example, communication interface 250 may include a network interface card (NIC) configured to communicate according to the Ethernet protocol. Additionally, communication interface 250 may implement a TCP/IP stack for communication according to the TCP/IP protocols. Various alternative or additional hardware or configurations for communication interface 250 will be apparent.

Storage 260 may include one or more machine-readable storage media such as read-only memory (ROM), random-access memory (RAM), magnetic disk storage media, optical storage media, flash-memory devices, or similar storage media. In various embodiments, storage 260 may store instructions for execution by processor 220 or data upon which processor 220 may operate. For example, storage 260 may store an operating system 261 for controlling various operations of system 200.

It will be apparent that various information described as stored in storage 260 may be additionally or alternatively stored in memory 230. In this respect, memory 230 may also be considered to constitute a storage device and storage 260 may be considered a memory. Various other arrangements will be apparent. Further, memory 230 and storage 260 may both be considered to be non-transitory machine-readable media. As used herein, the term non-transitory will be understood to exclude transitory signals but to include all forms of storage, including both volatile and non-volatile memories.

While system 200 is shown as including one of each described component, the various components may be duplicated in various embodiments. For example, processor 220 may include multiple microprocessors that are configured to independently execute the methods described herein or are configured to perform steps or subroutines of the methods described herein such that the multiple processors cooperate to achieve the functionality described herein. Further, where one or more components of system 200 is implemented in a cloud computing system, the various hardware components may belong to separate physical systems. For example, processor 220 may include a first processor in a first server and a second processor in a second server. Many other variations and configurations are possible.

According to an embodiment, storage 260 of system 200 may store one or more algorithms, modules, and/or instructions to carry out one or more functions or steps of the methods described or otherwise envisioned herein. For example, the system may comprise, among other instructions or data, ventilation instructions 262, waveform instructions 263, a trained asynchrony algorithm 264, and/or reporting instructions 265, among many other possible instructions and/or data.

According to an embodiment, ventilation instructions 262 direct the system to operate the ventilator device 270 of the ventilation system 200, specifically to ventilate a patient. The ventilator device can be any device suitable to ventilate a patient for any reason requiring ventilation. One or more parameters of the ventilator device instructions can be set, adjusted, preprogrammed, or otherwise determined by a healthcare professional. Thus, the ventilation instructions 262 enable operation of the ventilation device to ventilate the patient.

According to an embodiment, waveform instructions 263 direct the system to calculate an airway pressure water accumulation index from the received airway pressure waveform for a patient breath, and a flow water accumulation index from the received flow waveform for a patient breath. The ventilator system can calculate the airway pressure water accumulation index and the flow water accumulation index using a variety of different methods, and optionally calculates these indexes on a breath-by-breath basis. According to an embodiment, the ventilator system calculates the airway pressure water accumulation index and the flow water accumulation index using the methods described in conjunction with method