Personalized peritoneal dialysis treatment using design of experiment techniques
The described technology generally includes systems and processes for a PD optimization process may operate to estimate, predict, or otherwise determine the value of PD dose variables values based on patient characteristics and/or PD prescription information. In one embodiment, a PD optimization process may be or may include a UFV determination process, operative to determine a predicted UFV for a patient. In some embodiments, the UFV determination process may include training a computational model to generate a predicted UFV output based on input of patient characteristics, PD prescription information, PD outcomes (for instance, UFV), and/or historical information associated with patient characteristics, PD prescription information, and/or PD outcomes. In some embodiments, a feedback control process with continuous Intraperitoneal Pressure (IPP) and hydration status measurements may be used to keep the hydration status of the patient within a target level ran. Other embodiments are described.
1 . An apparatus, comprising:
at least one processor;
a memory coupled to the at least one processor, the memory comprising instructions that, when executed by the at least one processor, cause the at least one processor to:
access prescription information for a peritoneal dialysis (PD) process for a patient; and
provide the prescription information as at least one input to at least one computational model to determine a predicted intraperitoneal volume (IPV), wherein the at least one computational model is trained to determine the predicted IPV based on the prescription information and patient information for at least one prior PD process cycle for the patient or a population of patients associated with the patient;
wherein the prescription information includes a dwell time for the PD process and the patient information includes at least a hydration status of the patient, wherein the at least one processor is further caused to:
send a control signal to a controller of a PD machine to operate the PD machine according to the dwell time for the PD process for the patient, wherein the control signal is generated based on the hydration status of the patient, wherein the at least one computational model includes fuzzy logic configured to vary training data used to train the at least one computational model to determine the predicted IPV to minimize fluctuations of the hydration status of the patient around a target range.
2 . The apparatus of claim 1 , the prescription information comprising a fill volume and osmotic agent information;
wherein the at least one computational model is a machine learning or artificial intelligence model.
3 . The apparatus of claim 1 , wherein the patient information comprises at least one of activity level and treatment time-of-day.
4 . The apparatus of claim 3 , the instructions, when executed by the at least one processor, to cause the at least one processor to provide the patient information as the at least one input to the at least one computational model to determine the predicted IPV.
5 . The apparatus of claim 1 , wherein the controller of the PD machine comprises a proportional-integral-derivative (PID) controller, wherein the PID controller is configured to perform intraperitoneal pressure (IPP) and hydration status measurements on the patient to maintain the hydration status of the patient within the target range.
6 . The apparatus of claim 1 , wherein the at least one processor is further caused to:
receive a measured IPV during the PD treatment process for the patient;
determine an error value based on a difference between the predicted IPV and the measured IPV; and
update a training of the at least one computational model using the error value.
7 . The apparatus of claim 1 , wherein the at least one computational model is trained to determine a predicted ultrafiltration volume (UFV) based on the at least one input of the prescription information and a measured drain volume for the PD process.
8 . The apparatus of claim 7 , the instructions, when executed by the at least one processor, to cause the at least one processor to:
receive the measured drain volume for the PD process, and
provide the measured drain volume and the prescription information to the at least one computational model to determine the predicted UFV.
9 . The apparatus of claim 7 , the at least one computational model configured to determine the predicted UFV according to the following regression model:
UFV Pred ( t )= a+b ·[Glu]+ c·T dwell +d ·[Glu] 2 +e ·[Glu]· T dwell +f·T dwell 2 ,
where a, b, c, d, e, and f are regression model coefficients specific to the patient;
wherein Glu is a concentration of an osmotic agent and T dwell is a dwell duration.
10 . The apparatus of claim 7 , the instructions, when executed by the at least one processor, to cause the at least one processor to:
receive a measured UFV for the PD process,
determine an error value based on a comparison of the predicted UFV and the measured UFV, and
determine a health condition of the patient based on the error value.
11 . The apparatus of claim 10 , the health condition comprising peritonitis.
12 . An apparatus, comprising:
at least one processor; and
a memory coupled to the at least one processor, the memory comprising instructions that, when executed by the at least one processor, cause the at least one processor to:
receive prescription dosage information for a peritoneal dialysis (PD) process for a patient;
determine a predicted ultrafiltration volume (UFV) for the patient by processing the prescription dosage information via a trained computational model, wherein the trained computational model is trained based on at least one prior PD process cycle of the patient and a corresponding ultrafiltration volume measurement;
receive a measured UFV for the PD process of the patient;
determine an error value based on a comparison of the predicted UFV and the measured UFV;
determine a treatment recommendation for the patient based on the error value;
determine a hydration status of the patient; and
send a control signal to a controller of a PD machine to control a dwell time of the PD process based on the hydration status of the patient, wherein the PD process includes using the PD machine to treat the patient according to the treatment recommendation during the dwell time, wherein the trained computational model includes fuzzy logic configured to vary training data used to train the trained computational model to predict an IPV to minimize fluctuations of the hydration status of the patient around a target range.
13 . The apparatus of claim 12 , the prescription dosage information comprising a fill volume, a dwell time, and osmotic agent information;
wherein the trained computational model is a machine learning or artificial intelligence model.
14 . The apparatus of claim 12 , the instructions, when executed by the at least one processor, to cause the at least one processor to:
receive patient information; and
determine the predicted UFV using the trained computational model based on the patient information and the prescription dosage information.
15 . The apparatus of claim 14 , the patient information comprising at least one of time of day, activity level, or fluid intake.
16 . The apparatus of claim 12 , the instructions, when executed by the at least one processor, to cause the at least one processor to:
determine a health condition of the patient based on the error value.
17 . A method, comprising:
receiving prescription dosage information for a peritoneal dialysis (PD) process for a patient; and
determining a predicted ultrafiltration volume (UFV) for the patient by processing the prescription dosage information via a trained computational model, wherein the trained computational model is trained based on at least one prior PD process cycle of the patient and a corresponding ultrafiltration volume measurement;
receiving a measured UFV for the PD process of the patient;
determining an error value based on a comparison of the predicted UFV and the measured UFV;
determining a treatment recommendation for the patient based on the error value;
determining a hydration status of the patient; and
sending a control signal to a controller of a PD machine to control a dwell time of the PD process based on the hydration status of the patient, wherein the PD process includes using the PD machine to treat the patient according to the treatment recommendation during the dwell time, wherein the trained computational model includes fuzzy logic configured to vary training data used to train the trained computational model to predict an IPV to minimize fluctuations of the hydration status of the patient around a target range.
18 . The method of claim 17 , comprising determining a health condition of the patient based on the error value;
wherein the trained computational model is a trained machine learning or artificial intelligence model.
19 . The method of claim 17 , comprising performing a PD treatment on the patient based on the predicted UFV.
20 . The method of claim 17 , the prescription dosage information comprising a fill volume, a dwell time, and osmotic agent information.