NL2025657A - Smart inhalation drug delivery device - Google Patents

Abstract

In this patent, a smart inhalation drug delivery device aiming at optimizing systematic and local drug delivery. This device has the capability of regulating pressure drop in its conduits, according to a patient’s respiration characteristics and the embodied drug’s particle morphology. The device comprises a controlling system that adjusts the pressure drop by means of a valve Which is located in the inlet of the device’s drug delivery conduits. Patient’ s inspiration and opening factor of the valve profiles are modeled, predicted and regulated by machine learning methods and artificial intelligence. The drug delivery conduits of the device are designed to maximize the de-agglomeration of possible drug clogs while keeping pressure drop at a minimum. The replaceable aerosol source of the device loads the desired dosages with no need of gravity, either by means of a spring or compressed air force. In addition, this source benefits from sensors monitoring drug storage conditions. The device also provides a dosage adjustment capability where either a physician can remotely regulate the dosage, based on recommendations generated by an Al algorithm using health indices of a given patient through an IoT platform, or through a physician’s direct order and assessments.

Description

SMART INHALATION DRUG DELIVERY DEVICE Effective respiratory drug delivery depends strongly on a palient’s respiratory characteristics and behavior, device performance and drug particle morphology. Thus, inhalation drug delivery devices should be designed and personalized based on respiratory behavior and other influential parameters to improve drug delivery. Numerical simulations of respiratory drug delivery show that there is an optimum velocity for the entrance of drug particles into a patient’s month; this optimum velocity is related to the drug particle morphology [1-3]. Therefore, if the patient's respiration flow rate differs from the optimum flow rate, the desired drug delivery will not be obtained. As different patients have different respiratory characteristics, a limited number of patients fall into the optimum velocity domain for respiratory drug delivery. The disclosed device changes its pressure drop to obtain optimum flow rate and velocity at the mouth entrance with no need for the patient to change his/her respiratory behavior, The mentioned feature also helps the treatment of children or elderly with different respiratory patterns. It is also possible to monitor patient’s and drug’s conditions.

On the other hand, drug particles which are adjacent to each other, form aggregates of particles over time. This phenomenon causes a change in the morphology of particles and consequently leads to the reduction of drug delivery efficiency. So, air conduits of device should cause the de- agglomeration of aggregated drug particles by applying some techniques.

The presently disclosed subject matter provides a smart inhaler system to reach a complete personalized device and comprises an aerosel source, air conduits and entrance air flow rate control system, The entrance airflow rate control system is comprising of a flow meter sensor, a control valve, and a central processing unit. Flowmeter sensor is located at the mouthpiece § (FIG 1) and measures the airflow rate Q. Optimum flow rate Qo is calculated according to the particles morphology {i.e. shape, size and density), It should be noted that since drug particles have low Stokes number and follow fluid streamlines (perfect advection), the entering air velocity and the drug particles velocity are the same. The fresh air flow rate entering the air conduits is controlled with valve 8 (FIG 2). The flow rate Q is regulated to reach the optimum flow rate Qope. If ( goes above Om the opening factor K of the valve is reduced and the fresh air flow rate will decrease and vice versa. The value of K is in range 0 to 1, where 0 is for a completely closed valve and 1 is for a completely open valve. The relation between the Q and K is different for various patients, so, if is not possible to determine profile of Kop for Qop with a previously determined relation. To overcome this issue, the following algorithm is considered: during the use of the device, the flow rate of a patient (0) is continuously measured with the sensor and compared with Qope; then the value of the K is changed proportional to the difference between the Q and Qop and previous data set of inhalation experiences. To be more clear, to reduce K fluctuations which would results in reduction of Q fluctuation, K, is obtained from equation 1: (Ego. 1) Kot) = WOK, (0) + (1 WOE) Korat) where Ky is the real time valve opening factor at time t, K, is instantaneously calculated proportional to the difference between the Q and Op and Kaw Is Ky in previous inhalation, W is the weight function and determines how much the K, is effected by previous Ko. Wii) is proportional to AQ {1} /Q ave, where AQ {£) is the difference between Q(t) and Qave(t) and Que 18 average of Qt) in all previous inhalations, The algorithm leads to using Koe when there is no significant changes in patient respiratory profile between inhalations to minimize UQ fluctuations. When the inspiration profile differs from average profiles due to any factors like health conditions, proportional to amount of this difference, K, is determined more by K, to more effectively correct Q.

This procedure is indicated schematically and briefly in FIG 10. The obtained pattern will become more precise each time and the optimum and desired flow rate will be reached with less fluctuations.

The pattern obtained from the respiratory behavior of the various patients will be evolved over time. This helps the faster pattern-finding for the other patients. To do this the following approach is proposed: During the usage of the device, a set of information about the patient is gathered through different ways: 1) Personal information, provided by the individual himself/herself such as weight, height, age, sex, job, etc. 2) Medical records such as previous disease records, blood pressure, blood sugar, etc. 3) Information gathered by the product and during the usage of the device using sensors 3¢ embedded in the product. This information is saved in a database and after the data cleaning and preprocessing step, machine learning algorithms will be applied, to learn and recognize existing patterns in the data and use the patterns to make predictions of future outcomes.

This process is called supervised learning.

In other words, supervised learning is the process of mapping a set of input data, i.e, independent variables to their target output, ie. the dependent variable, In this case, input data consists of the set of information and features, describing each patient and the corresponding output is the optimum value of K for that patient.

When the training phase is finished, the value of the opening factor for different patients considering their features is learned.

Finally, the {rained algorithms can be used to make predictions for new patients.

To be more specific, when a new patient is using the device, the features of that patient would be given to the trained algorithm, and the output, ie.

Kog, would be generated almost immediately.

Since the output is a real-valued scalar, this problem is considered as a regression problem.

Table 2 presents an example of a dataset containing features of patients, considered as inputs and their corresponding output.

This dataset should be divided into training and test set.

Then a machine learning algorithm, in particular, a deep learning algorithm is applied to the {raining set 18 to learn the complex relationship between input features and their corresponding output value, Deep learning methods have been successfully applied to many real-world tasks, including medical problems{4]. Deep learning methods use deep artificial neural networks which consist of an input layer, an output layer and a variable number of hidden layers.

Each layer consists of some nodes, which are connected to the next layer’s nodes.

Each edge connecting two nodes is assigned with a weight that determined the strength of the connection between the node pair.

The goal of deep learning algorithms is to learn these weights such that the trained network minimizes the prediction error, After the training step is done, the test data is feed into the trained network to generate predictions, s0 we can evaluate the performance of our algorithm.

In FIG 11 an example of a deep neural network is provided.

The appropriate network architecture for this specific problem can be obtained in the experimental design process.

Sensors for measuring respiratory behavior send their data to a cloud system in real-time.

This data is then pre-processed with suitable algorithms to be ready for training and evaluating algorithms.

In other words, device takes advantages of the technologies based on IoT' to better integrate and utilize collected data from different sources, Furthermore, io control this device an ì Internet of Things application is installed on the patient's smartphone to gather the information through this application and by the patient.

In the disclosed subject matter a hollow body as drug delivery conduits is presented.

The main effective parameters on the de-agglomeration of the drug particles are shear stress exerted from the ambient fluid (air), turbulence, the collision of the drug particles to the walls and the interaction of the particles with each other.

On the other hand, the pressure drop plays an important role, because the increase of the pressure drop causes a lower airflow rate applied by the patients with weaker respiration ability and consequently reduces the effectiveness of drug delivery.

It means that the de-agglomeration factors should not cause the growth of pressure drop; so, there 19 is an optimum point.

Thus, in the presently disclosed subject matter using simulations carried out with CFPD?, a unique air conduit (ie. intertwined curved zigzag) is offered.

Also, in this geometry, parameters such as conduits diameter, curvature radius, path length, and deviation angles are optimized.

This optimization is done by simulating 4 different geometries (i.e. simple circular cross-section tube, FIG 9A, helical tube, FIG 9B, zigzag, FIG 8C and curved zigzag, FIG 9D) for air conduits.

Three important parameters {averaged wall shear stress, pressure drop and volumetric averaged turbulent viscosity) are compared in these geometries and the obtained results are presented in Table 1. Wall shear stress and turbulent intensity are desired and pressure drop is undesirable.

It can be seen in Table 1 that wall shear stress and turbulent viscosity {as a criterion for turbulent intensity) in both zigzag shape geometries are bigger than the ones in the two other shapes.

But, in curved zigzag, the pressure drop is lower than the simple zigzag.

Based on these results, curved zigzag is selected among the other shapes.

Then, to create the pressure drop alteration and enhancement of particle-fluid interaction due to the secondary flow, the final unique shape consisting of two curved zigzags is proposed.

This geometry results in desired shear stress, turbulence and the drug particle collision with walls and with each other, while minimizes the pressure drop.

Simulation results show that controlled fresh air entrance enhances the creation of secondary flows, turbulent intensity and the other parameters affecting the drug particles de-agglomeration (see FIGs. 184, 19B and 19C in which drug particles velocity, turbulent intensity, and wall shear stress distribution are shown respectively); hence, the operation of air conduits is improved and with the increase of 2 Computational Fhid-Particles Dynamics interaction of the two flows (i.e. airflow containing the drug particles and controlled fresh air) this improvement intensifies, Regarding FIG 8, air conduits include two separate inlets for the entrance of the air containing drug particles 24 and the fresh controlled air 25 and an outlet 23, Air containing drug particles is 5 supplied by feeding and dose regulating system and the other airflow is supplied by a valve 11. Air conduits have unique shape 27. The hollow body 26 is rectangular and placed in the air routes compartment in the main body of the device.

In the disclosed subject matter, the aerosol source is capsule based, reservoir based and blister based source.

In the feeding part, the required amount of the drug is saved in reservoir for a determined time.

To make the operation of this part independent from the direction of the device placement, at least one spring and a plate 37 are used (FIG 4). This mechanism makes it possible to use the device on the horizontal state or even in the direction opposing to the earth gravity direction.

This plate 37 is placed between the remaining drug and the reservoir cap 31. In another disclosed subject matter, compressed air is used between the plate and reservoir cap.

In another disclosed subject matter, aerosol source is replaceable.

The presently disclosed subject matter provides a smart inhaler device in which the drog condition is monitored by the means of a thermometer sensor and a psychrometer sensor and physician and patient are informed of any possible damages.

If automatic mode is activated, the damaged drug administration would not be possible.

Dose feeding and regulating system (FIGS 4 and 5) is comprising of two main parts: feeding part 33 and regulating part 28-29, 34-36. The feeding part supplies the required drug and the drug regulating part transfers the required dose into the drug delivery conduits.

After supplying the drug by the drug feeding part, the drug is placed adjacent to the drug dose regulating 29 perforations (FIG 6). Then, by rotation of the disc 36 and with the help of the excrescences on the walls 35 (FIG 7) the drug is guided in three stages on the perforations and located in the perforations by paddles 34. Then, regarding the amount of using dose, a determined number of perforations are located on the drug delivery conduits 32. The size of the perforations is designed in a way that after the placement, the drug particles do not fall due to gravity and remain in the perforations.

These two parts are connected with a pin 28 and pinhead

30 (FIGS $, 6). As said before, the using dose is controlled by the number of involved perforations. Perforations are designed to be positioned angularly on the dose regulating disc and the number of perforations that should be located in drug delivery conduits can be determined by controlling the rotation angle with an electromotor which is connected to the 5 regulating part. Since by increasing the number of perforations, the number of dosages which can be regulated would also increase, the cross-section area of the perforations is designated to have maximum number of dosages available. Also, this shape of the perforations of cross section is suitable for drug particles exhaust and the drug particles can be entered into the conduits with minimum airflow rate.

Drug dose determination is always out of the patient’s control and should be done with the final confirmation of the physician. In some cases, by monitoring disease symptoms {e.g. monitoring the blood pressure in patients with hypertension or monitoring blood sugar level in diabetic patients) or patient’s health indices, the device recommends a change in the drug dosage io the physician through an online application. Then, after the confirmation of the physician, the dosage change order will be given to the device and the drug dose can be changed online. Also, the physician can change the drug dose based on his/her assessment.

In the case of running out of the drug, the same approach can be applied, The design of the device is done in a way that the patient can change the finished or damaged drug alone. All the mentioned parts are located in the main body of the device 1 (FIG 1). Power button 2, LCD 3 and preparing button 4 are assembled in 19, 18 and 17, respectively. Furthermore, there are curvatures 28 on the body of the device for easy use and comfortable hand-holding, Electricity required is supplied by a battery which is located in 13. This battery connects with the electronic boards placed in 15, 16 and electromotor placed in 14.

Dose feeding and regulating system are placed in 19 where two conduits are beneath this system and above it. Inlet air to the dose regulating perforations 21 and the location of the electromotor shaft 22 is beneath of it; above this there are drug particles containing air conduit 13 to drug delivery conduits. In the main body, an air inlet 9 is also placed to let the fresh air enter into drug delivery conduits through a pathway 7 after passing the flow rate control valve.

Another aspect considered for this device is the encouragement of the patient to breathe-holding.

Immediately after the finishing of drug's breathing, the device starts to count the time and encourages the patient to hold his/her breathe up to the end of the counting.

This time can be controlled and changed online by a physician considering the patient’s condition.

It is also mentioned that the presently disclosed subject matter provides a smart inhaler device that can monitor the patient's condition.

So, the physician can check the time and the quality of the drug consumption by the patient; as this device can measure the flow rate and compare it with the optimum flow rate.

The physician can also put the device in a limited mode; in this mode, the preparation of the drug by the device will be activated in determined times and the patient cannot use the device outside these times.

Furthermore, the physician with the help of considering reported symptoms can order the use of the drog with a suitable dose.

In this state,

the device can announce the drug usage command to the patient with the help of its tools such as a speaker, LCD, and the patient’s smartphone.

Claims (1)

2 ~ 6 Conclusions What is claimed is: 1. An inhalation drug delivery device 1 with pressure drop control of the tubing of the device, according to a patient's respiratory characteristics and drug particle morphology, to optimize systematic and local drug atyifie. This device consists of: a. An acrosol source 33, b. a dose loader 36 that transfers the drug from the aerosol source to the drug delivery lines 26 of the device, and c. a hollow body 26 as a path for at least one fluid to flow to transfer the drug from dose charger 36 to mouthpiece 3, comprising at least one inlet 24 of dose loader 36 and at least one inlet 25 of fresh air and at least one inlet 25 of fresh air. one outlet 23, the pressure drop in the hollow body being controlled by fresh air flow. d. control system consisting of a flow rate sensor and a valve and a pmocessor, where the processor controls the pressure drop of the valve according to the difference in measured and optimal flow rates. A drug delivery device by inhalation according to claim 1, wherein the pressure drop control mechanism is controlled by a "machine learning" method and artificial intelligence. A drug delivery device by inhalation according to claim 1, wherein the hollow body has a plurality of curvatures 27 to minimize possible agglomeration of drug particles. 4. Weight for the drug delivery device by inhalation according to claim 1, wherein the aerosol source is a capsule-based, reservoir-based and blister-based source. 5, Drug delivery device by inhalation according to claim 4, wherein the aerosol source of the reservoir consists of a plate 37 and a plurality of springs, the drug being loaded in a plurality of perforations on the plate using a compressive force, 6, Drug delivery device by means of inhalation according to claim 4, wherein the aerosol source of the reservoir consists of a plate 37 and compressed air, the medicament being loaded onto the plate in a plurality of perforations by means of a compressive force, An inhalation drug delivery device according to claim 1, wherein the acrosol source is replaceable. The inhalation drug delivery device of claim 4, wherein the aerosol source of the reservoir is at least one sensor to monitor the drug condition. 3, Drug delivery device by inhalation with drug adjustment, which consists of: a. A number of perforations 29 of the dose loader 36, and a control mechanism to determine the number of perforations of the dose loader to be placed on the path of the cavernous body control to achieve the required dose, identifying the required dose by machine learning techniques, using a number of sensors to collect a data set with patient health indexes as input and predict their corresponding output in real time , A drug delivery device by inhalation according to claim 9, wherein the control mechanism uses a physician's directions to control the number of dosing loader perforations that must be placed in the path of the drug delivery conduits to achieve the required dose,