Personalized modeling for drug concentration prediction using Support Vector Machine

Abstract

Abstract-Building a personalized model to describe the drug concentration inside the human body for each patient is highly important to the clinical practice and demanding to the modeling tools. Instead of using traditional explicit methods, in this paper we propose a machine learning approach to describe the relation between the drug concentration and patients' features. Machine learning has been largely applied to analyze data in various domains, but it is still new to personalized medicine, especially dose individualization. We focus mainly on the prediction of the drug concentrations as well as the analysis of different features' influence. Models are built based on Support Vector Machine and the prediction results are compared with the traditional analytical models.

Safe Implementation of Embedded Software for a Portable Device Supporting Drug Administration

“…Several personalized drug concentration prediction method based on Support Vector Machine (SVM) algorithm where presented in our prior works [20]- [22]. The initial method was only able to perform a point-wise drug concentration prediction, therefore, it is impossible to calibrate in personalized manner the prediction every time when a new measured concentration value is available for the patient under treatment.…”

“…for whom this measurement was performed. To build the analytical representation of the DCT curve, ParaSVM uses the common basis functions β j = {t −2 , log(t), 1 − e −t }, respecting the shape of DCT curve obtained from the PK method [2], where t stands for time [22]. Therefore, the target is to obtain the parameters y for the weights of β :…”

“…Unlike in [20] and [26] where the SVM algorithm was used to predict the drug concentration values based on patient parameters, here the SVM algorithm is applied to learn the mathematical relationship between the parameters of the basis functions and then to predict the parameter values of the DCT curve for a new patient in the testing dataset. With one given measured concentration value, in case of a posteriori adaptation, or a value of the residual concentration after the previous dose intake, the curve parameters can be adjusted, this way allowing to build the personalized DCT curve for each next cycle.…”

Parameterized SVM for personalized drug concentration prediction

“…In addition, the influence of different patients' features on the accuracy of predicting the drug concentrations is also illustrated. In [5] an extension of the prediction algorithm by combining the RANdom SAmple Consensus (RANSAC) algorithm with SVM was proposed. This extension has improved the prediction accuracy by about 40%.…”

“…The parameter library of the basis functions used for curve interpolation are computed using the RANSAC algorithm previously applied in [5] to separate inliers and outliers from a set of (noisy) data. Using this method, we are able to process as many features as possible, consider binary inputs, visualize the data represented in a form of personalized DCT curve for each patient, and adjust it structurally to make it more personalized with a given measured concentration value.…”