Insulin Recommender - Overview

Authors: Ferran Torrent-Fontbona, Beatriz López and Joaquim Massana (Univeristy of Girona).

1. Introduction

The purpose of this document is to describe the setup procedure for the PEPPER Insulin Recommender library, the insulin recommender API and the details on the implementation. The PEPPER Core library (Insulin Recommender) is a self-contained independent library designed for interaction with the PEPPER Safety System. The core of the library and the API have been developed by the University of Girona.

2. PEPPER core library setup

This chapter details the setup of the PEPPER Insulin Recommender library.

2.1. Importing the PEPPERCore.jar

Setup of the library involves only importing the PepperCore.jar into the Android project. In this section we provide two options: First, manually edit the application’s module build.gradle; Second, using the Android Studio interface to import a JAR as a dependency. The PepperCore.jar can be downloaded from this link:

http:// pepper.eu.com/ApplicationFiles/Download/791

2.1.1. Option 1: build.gradle

1. Place the PepperCore.jar in the libs folder of the project, or a direction of your preference.
2. You now have two options:
   • Option 1: Ensure that the directory and its JARs are compiled by checking/adding the following dependency in the application module’s build.gradle (example for the libs directory):

     
     compile fileTree(include: [’*.jar’], dir: ’libs’)
     
     

     For example:

     
     dependencies { compile fileTree(include: [’*.jar’], dir: ’libs’)
     // other dependencies omitted
     }
     
     

   • Option 2: Add an explicit dependency for the PepperCore.jar to the application module’s build.gradle (example for the libs directory):

     
     compile files(’libs/PepperCore.jar’)
     
     

     For example:

     
     dependencies { compile files(’libs/PepperCore.jar’)
     // other dependencies omitted
     }
     
     

3. After the build.gradle has been sync’d, the library should not be imported.

2.1.2. Option 2: Android Studio Graphical User Interface

1. Place the PepperCore.jar in the libs folder of the project.
2. Navigate to File > Project Structure.
3. Select your module from the Modules list in the left side pane of the Project Structure interface.
4. Navigate to the Dependency tab.
5. Click the add icon (+ symbol) and select File dependency.
6. Navigate to the PepperCore.jar and click OK.
7. Click OK to close the Project Structure window.
8. The library should now be imported.

3. CBR bolus recommender system architecture

The architecture of the CBR System inside a CBR core. This core holds the CBR classes within three main packages: CBR, IO and PEPPER:

3.1. Core.CBR

Location:.\src\Core\CBR 

This package contains the CBR class, the main class for cbr execution that calls all other classes to perform the cycle. The package also contains the following data and method packages needed for the execution:

3.1.1. Distances

Location:.\src\Core\CBR\Distances 

This package contains the classes to compute distances between Populations, its Individuals and their Attributes.

Globals 
Location:.\src\Core\Distances\Globals 

This package contains the classes to compute different kinds of distance between all the Individuals of the train Population towards the Individuals in the test Population, considering all of their Attributes. These classes extend the abstract class GlobalDistance, and are to be built through the GlobalDistanceFactory class.

Locals 
Location:.\src\Core\Distances\Locals 

This package contains the classes to compute different kinds of distance between all the Individuals of the train Population towards the Individuals in the test Population, considering only one specific Attribute. The LocalDistance class used for an Attribute should match its signature (i.e. don’t use LocalDistanceEuclidean for text Attributes). These classes extend the abstract class LocalDistance, and are to be built through the LocalDistanceFactory class.

Unknown 
Location:.\src\Core\Distances\Unknown 

This package contains the classes to compute different kinds of distance between all the Individuals of the train Population towards the Individuals in the test Population, considering only one specific Attribute, for which at least one of said Individuals lacks its value. These classes extend the abstract class Missing, and are to be built through the MissingFactory class.

3.1.2. Models

Location:.\src\Core\CBR\Models 

This package contains the data structures to hold the necessary information for the multiple computations of the cbr cycle. The basic, most used classes in this package are Attribute, Individual and Population.

 

Attribute 

This class defines the properties of one of the multiple features of each case. It holds its identifier, its description, its type, and its weight.

Individual
 

This class defines a case. It holds an identifier, and the values of all its features. It also references the Population to which the Individual belongs.

 

Population 

This class defines a case base. It holds all of its Individuals and a collection of Attributes, indicating which features its Individuals hold, and its type. This package also contains, in the Reuse inner package, the data structures to hold the outputs of the said computations.

Results 
Location:.\src\Core\Models\Results 

This package contains the classes needed to hold the data resulting from the cbr computations, such as distance matrixes, Population classifications, selections, etc.

3.1.3. Retrieve

Location:.\src\Core\CBR\Retrieve 

Contains both the context reasoning and case selection methods.

Context filtering 
Location:.\src\Core\CBR\Retrieve\ContextReasoning 

The classes in this package are built for the purpose of filtering cases from a Population, keeping only those with features than meet specific criteria. These classes extend the abstract class ContextReasoning, and are to be built through the ContextReasnongFactory class.

Selection 
Location:.\src\Core\CBR\Retrieve\Selection
 

The classes in this package are built for the purpose of selecting similar train cases (from the CBR’s train Population) for every test case (from the CBR’s test Population). Each class may select a different number of cases meeting different specific criteria. These classes extend the abstract class Selection and are to be built through the SelectionFactory class.

3.1.4. Reuse

Location:.\src\Core\CBR\Reuse 

The classes in this package are built for the purpose of classifying every test case considering the data of the train cases selected for it in the Retrieve phase. These classes extend the abstract class Reuse and are to be built through the ReuseFactory class.

3.1.5. Revise

 

Location:.\src\Core\CBR\Revise 

The classes in this package are built for the purpose of checking whether the classification made in the Reuse phase is correct, and make any possible corrections to each misclassified case. These classes extend the abstract class Revise and are to be built through the ReviseFactory class.

3.1.6. Retain

Location:.\src\Core\CBR\Retain 

The classes in this package are built for the purpose of deciding which already classified test cases can be useful as train ones, and add those meeting the criteria to the train Population. These classes extend the abstract class Retain and are to be built through the RetainFactory class.

3.1.7. Review

Location:.\src\Core\CBR\Review 

The classes in this package are built for the purpose of deciding which cases from the train Population are truly relevant and which can be dropped following specific criteria, in order to keep the case base as small as possible without losing relevant information. These classes extend the abstract class Review and are to be built through the ReviewFactory class. It includes the restore procedures.

3.1.8. Exceptions

Location:.\src\Core\CBR\Exceptions 

This package contains exception classes to clarify cbr execution errors and pinpoint them easily whenever they appear.

3.2. Core.BaseLearners

Location .\src\Core\BaseLearners 

This package contains the classes intended for machine learning execution over the weights given to the features of the cases.

3.3. Core.Distributed Approaches

Location:.\src\Core\DistributedApproaches

This package contains the classes built for the purpose of executing cbr functionalities in a distributed way. The implemented distributed functionalities are both the cbr cycles, in MultiAgent, and the weight learning phase, in MetaCBR:

 

MultiAgent 
Location:.\src\Core\DistributedApproaches\MultiAgent 

Package which classes are intended for the distributed execution of multiple cbr cycles.

MetaCBR 
Location:.\src\Core\DistributedApproaches\MultiAgent 

Package which classes are intended for the distributed execution of a generic algorithm to learn the best Attribute weights for the cbr Populations.

3.4. Core.Tools

Location:.\src\Core\Tools 

This package holds multiple classes intended to encapsulate and reuse code of basic functionalities used throughout the whole of the cbr execution, such as statistics computation, string manipulation, etc.

3.5. IO

 
Location:.\src\Core\IO

The IO package contains the classes for saving and reading data in CSV or JSON formats.

3.5.1. Access2CSV

Location:.\src\Core\IO\Access2CSV

3.5.2. Exceptions

Location:.\src\Core\IO\Exceptions 

This package contains exception classes to clarify in and out operation errors and pinpoint them easily whenever they appear.

3.5.3. JSON

Location:.\src\Core\IO\JSON 

This package holds the Google’s Gson library for JSON construction and reading. It also holds the custom Jsoner class, which for a given class allows to transform its objects to a JSON string and from JSON a String instantiates the original class object.

3.5.4. PEPPER

Location:.\src\Core\IO\PEPPER

This package holds the classes to allow building and reading of the PEPPER standard JSONS from the cbr data. This is needed because the Gson package needs an actual object with the exact instantiation the JSON is describing in order to make this JSONS and create objects form them. It also contains a Manager class, which encapsulates the multiple Jsoners for all needed data classes and has methods to build the said PEPPER standard JSONS and JSON objects.

3.6. PEPPER

This package holds the classes intended to preprocess incoming data into the format intended for the cbr execution. For instance, it holds the PhysicalActivity class, which quantifies steps into PhysicalActivity intensity.

4. Using the Pepper CBR library

4.1. CBR setting

The first step is to set the parameters of the CBR. The easiest way to define the parameters is to use an interpreter such as the PepperInterpret (Section 5.2). The following instructions create the interpret and define the patient profile:

InterpretPepper interpret = new InterpretPepper();
interpret.setPatientData(patientGlow,patientGhigh,patientGsetpoint,
  patientCHOmean,patientWeight);

4.2. Building CBR

The next instruction builds a CBR with train and test populations:

CBR cbr = interpret.construeixCBR(test,train);

These populations can both be null, and set afterwards through the methods setTrain(Population) and setTest(Population). The train and test populations will be built from JSON objects, seen in Section 5.3. The train population should contain the list of individuals from the case base of the given user, while the test population will contain only one individual, that is a CBR JSON object.

The train population will be built from a full population JSON object:

Population train = Manager.populationFromJSON(trainJSON);
cbr.setTrain(train);

The test population should be conformed of just one individual, that is the current case of the patient, built from a JSON CBR object that should be filled with the corresponding patient data. To build a new population with just one individual, the JSON Manager will require an attributes JSON, seen in Section 5.4, with the list of attributes of the individual:

Population test = new Population();
test.setAttributes( Manager.attributesFromJSON( attributesJSON ) );
test.addIndividual( Manager.individualFromJSON( testIndividualJSON ) );

The required values to be initialised in the JSON CBR object, seen in Section 5.3, are:

PatiendId
CaseId
datetime
timeofday
typeofday
PremealBloodGlucose
PastPhysicalActivityIntensity
PastPhysicalActivityType
FuturePhysicalActivityIntensity
FuturePhysicalActivityType
BasalDose
InsulinOnBoard 
CarbsOnBoard 
Carbohydrates
Fats
AlcoholType
AlcoholQuantity
HoursOfSleep
PsychologicalStress
Happiness
Tiredness
HormoneCycle
Fever
DigestiveIllness
Medication
AmbientTemperature

4.3. Retrieve step

Execute the Retrieve step with this instruction:

cbr.retrievePepper();

4.4. Reuse step

Execute the Reuse step with this instruction:

cbr.reusePepper();

This steps fill the ICRreuse, ISFreuse and BolusReuse attributes in the case. Convert the full test population to JSON, or alternatively, convert just the current case to JSON and send it to the server if it is necessary:

String testPopuJSON = Manager.toJSON( cbr.getTest() );
String testCaseJSON = Manager.toJSON( cbr.getTest().getIndividual(0) );

Before to perform this step, the following attributes should be filled:

• ICRsafety: ICR (insulin to carbs ratio) value returned by safety system
• ISFsafety: ISF (insulin sensitivity factor) value returned by the safety system
• BolusSafety: Bolus value returned by the safety system
• ICRuser: ICR value calculated from the bolus administered by the user (Section 5.1.4)
• ISFuser: ISF value calculated from the bolus administered by the user (Section 5.1.2)
• BolusUser: final bolus administered by the user
• PostprandialMinGlucose: minimum value of the postprandial blood glucose. It can be calculated from SafetyObject.CGM[].MeasuredValue of the JSON safety system object.

4.5. Revise step

The following code executes the Revise step to correct the solution and fill the revise values ICRrevise (Section 5.1.1), BolusRevise (Section 5.1.3), ISFrevise (Section 5.1.2):

cbr.revise();

This is the instant in which traceability of the patient can be stored. It can be done saving the JSONs of the test, train and attribute data in one or separated files:

serverTrain = Manager.toJSON( cbr.getTrain() );
serverTest = Manager.toJSON( cbr.getTest() );
serverAtt = Maneger.toJSON( cbr.getTrain().getAttributes() );

4.6. Retain step

Execute the Retain step to add the new case to the case base:

cbr.retain();

4.7. Review step

The Review phase remove all the unnecessary cases:

cbr.review();

4.8. Learning weight

Execute the weight Learning phase:

cbr.weightLearning();

5. Annex

5.1. Equations

On the present section some of the used equations are explained.

5.1.1. Revision equation.

The following equation is used in order to calculate the ICR revised during the revision step.

$\begin{array}{ccc}& IC{R}_Q\text{'}=(1-\alpha )IC{R}_Q+\alpha \genfrac{}{}{1px}{}{CHO+\genfrac{}{}{1px}{}{G_c-G_{sp}}{341.94/W}}{B_Q+IoB+\genfrac{}{}{1px}{}{G_{\min }-G_l}{IS{F}_Q}}\hfill & \hfill \text{(1)}\end{array}$

On where:

α

Learning weight.

$IC{R}_Q$

Insulin to carbohydrates ratio used.

$CHO$

Carbohydrates intake.

$G_C$

Blood glucose value.

$G_{sp}$

Glucose set point (130).

$IoB$

Insulin on board.

$W$

Patient’s weight.

$G_{\min }$

Post prandial blood glucose.

$G_l$

Low glucose set point. 90 if it not exists.

$B_Q$

Bolus.

${ISF}_Q$

Insulin sensitivity factor used.

5.1.2. ISF equation.

The following equation is used in order to calculate the Insulin sensitivity factor (ISF).

$\begin{array}{ccc}& ISF=\genfrac{}{}{1px}{}{341.94\times ICR}{W}\hfill & \hfill \text{(2)}\end{array}$

On where:

$ICR$

Insulin to carbohydrates ratio used.

$W$

Patient’s weight.

5.1.3. Bolus reuse equation.

The following equation is used in order to calculate the bolus in the reuse step.

$\begin{array}{ccc}& B=\genfrac{}{}{1px}{}{CHO}{ICR}+\genfrac{}{}{1px}{}{G_C-G_{sp}}{ISF}-IoB\hfill & \hfill \text{(3)}\end{array}$

On where:

$ICR$

Insulin to carbohydrates ratio used.

$CHO$

Carbohydrates intake.

$G_c$

Blood glucose value.

$G_{sp}$

Glucose set point (130).

$IoB$

Insulin on board.

5.1.4. ICR equation.

The following equation is used in order to calculate the ICR on several steps.

$\begin{array}{ccc}& ICR=\genfrac{}{}{1px}{}{CHO}{B+IoB+\genfrac{}{}{1px}{}{G_{sp}-G_c}{ISF}}\hfill & \hfill \text{(4)}\end{array}$

On where:

$CHO$

Carbohydrates intake.

$G_c$

Blood glucose value.

$G_{sp}$

Glucose set point (130).

$IoB$

Insulin on board.

$B$

Bolus.

$ISF$

Insulin sensitivity factor used.

5.2. Interpret

The following code is need in order to chose the methods there is the intention to use:

package pepperapp;

import Core.CBR.Retrieve.Selection.Selection;
import Core.CBR.Retrieve.ContextReasoning.ContextReasoning;
import Core.CBR.Distances.Locals.LocalDistance;
import Core.CBR.Distances.Globals.GlobalDistance;
import Core.CBR.CBR;
import Core.CBR.Exceptions.FactoryException;
import Core.CBR.Models.*;
import Core.CBR.Reuse.*;
import Core.CBR.Revise.*;
import Core.CBR.Retain.*;
import Core.CBR.Review.*;
import Core.BaseLearners.LearningAttributeWeights.LearnNot;
import Core.BaseLearners.LearningAttributeWeights.LearningAttributeWeights;
import Core.CBR.Distances.Globals.*;
import Core.CBR.Distances.Locals.*;
import Core.CBR.Distances.Unknown.*;
import Core.CBR.Retrieve.ContextReasoning.FilteringByContext;
import Core.CBR.Retrieve.Selection.SelectionNK;

public class InterpretPepper{

    // Default patient data, must be set to real values before revise step;
    double Glow = 5;
    double Ghigh = 7;
    double Gsetpoint = 6;
    double CHOmean = 70;
    double weight = 70;

    public InterpretPepper()
    {
    }

    public CBR construeixCBR(Population test, Population train) throws FactoryException {
        
        /** Distance computing objects */
        Missing desconeguts = new SkipMissingFromTest();
        GlobalDistance dist_glob = new GlobalDistanceWM();
        LocalDistance dist_loc_num = new LocalDistanceEuclidean();
        dist_loc_num.setMissingMethod(desconeguts);
        LocalDistance dist_loc_disc = new LocalDistanceEuclidean();
        dist_loc_disc.setMissingMethod(desconeguts);
        LocalDistance dist_loc_text = new LocalDistanceBoolean();
        dist_loc_text.setMissingMethod(desconeguts);
        LocalDistance dist_loc_data = new LocalDistanceDateStandard();
        dist_loc_data.setMissingMethod(desconeguts);
        LocalDistance dist_loc_seq = new LocalDistanceSequencesDIST4MADCAPPablo();
        dist_loc_seq.setMissingMethod(desconeguts);
        
        dist_glob.setLocalDistances(dist_loc_num, dist_loc_disc, dist_loc_text, dist_loc_data, dist_loc_seq);
        
        /** CBR phases objects */


        Selection sel = new SelectionNK();sel.addParameters("3");
       // Reuse reuse = new //ReuseMajoriaK();reuse.afegeixParametres("1");

Reuse reuse = new ReuseNumericSimilarityBased();reuse.afegeixParametres("3");

        Revise revise = new RevisePepper(); ((RevisePepper)revise).setPatientData(Glow, Ghigh, Gsetpoint, CHOmean, weight);
        Retain retain = new TemporalAdaptiveRetain();
        Review review = new ReviewPepper_CRR();
        
        /** Medicines, illnesses, fever and hormone cycle have to match **/
        ContextReasoning filter = new FilteringByContext();
        filter.addParameters(/*"Drugs;=;"+*/"digestiveIllness;=;"+"fever;=;"+"hormonecycle;=");

LearningAttributeWeights learn = new SalzbergModified("ICRrevise");

     
        
        CBR cbr = new CBR(train, test, dist_glob, sel, reuse, revise, retain, "ICRrevise");
        cbr.afegeixReview(review);
        cbr.afegeixLearn(learn);
        cbr.afegeixFiltrat(filter);
        
        return cbr;
    }

    public void setPatientData( double Glow, double Ghigh, double Gsetpoint, double CHOmean, double weight ){
        this.Glow = Glow;
        this.Ghigh = Ghigh;
        this.Gsetpoint = Gsetpoint;
        this.CHOmean = CHOmean;
        this.weight = weight;
    }
}

5.3. Case example

Description: this section describes how the case file is composed.

[
{
"attributes": {
"PatientId": "null",
"CaseId": "null",
"datetime": " null",
"timeofday": "null",
"typeofday": null,
"PremealBloodGlucose": "null",
"UserPastPhysicalActivityIntensity": "null ",
"UserPastPhysicalActivityType": " null ",
"PastPhysicalActivityIntensity": " null",
"PastPhysicalActivityType": "null",
"FuturePhysicalActivityIntensity": "null",
"FuturePhysicalActivityType": "null",
"BasalDose": "null",
"InsulinOnBoard": "null",
"CarbsOnBoard": "null",
"Carbohydrates": "null",
"Fats": "null",
"MealAbsorptionRate": "null",
"AlcoholType": "null",
"AlcoholQuantity": “null",
"HoursOfSleep": "null",
"PsychologicalStress": null,
"Happiness": "null",
"Tiredness": "null",
"HormoneCycle": "null",
"Fever": "null",
"DigestiveIllness": "null",
"Medication": null,
"AmbientTemperature": "null",
"ICRreuse": "null",
"ISFreuse": "null",
"BolusReuse": "null",
"TypeOfBolus": "null",
"ICRsafety": "null",
"ISFsafety": "null",
"BolusSafety": "null",
"ICRuser": "null",
"ISFuser": "null",
"BolusUser": "null",
"PostprandialMinGlucose": "null",
"ICRrevise": "null",
"ISFrevise": "null",
"BolusRevise": "null",
"ExpertValidation": "null",
"RejectCase": "null",
"ReusedCases": "878718a6-cdb8-4757-95a4-9b593cb6de3f"
},
"id": "null"
}
]

5.4. Attributes example

Description: this section describes how the attributes file is composed.

 [ 
{ 
"id": "patientid", 
"description": "patientid", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "caseid", 
"description": "caseid", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "datetime", 
"description": "datetime", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "timeofday", 
"description": "timeofday", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "typeofday", 
"description": "typeofday", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "premealbloodglucose", 
"description": "premealbloodglucose", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "userpastphysicalactivityintensity", 
"description": "userpastphysicalactivityintensity", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "userpastphysicalactivitytype", 
"description": "userpastphysicalactivitytype", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "pastphysicalactivityintensity", 
"description": "pastphysicalactivityintensity", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "pastphysicalactivitytype", 
"description": "pastphysicalactivitytype", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "futurephysicalactivityintensity", 
"description": "futurephysicalactivityintensity", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "futurephysicalactivitytype", 
"description": "futurephysicalactivitytype", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "basaldose", 
"description": "basaldose", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "insulinonboard", 
"description": "insulinonboard", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "carbsonboard", 
"description": "carbsonboard", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "carbohydrates", 
"description": "carbohydrates", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "fats", 
"description": "fats", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "mealabsorptionrate", 
"description": "mealabsorptionrate", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "alcoholtype", 
"description": "alcoholtype", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "alcoholquantity", 
"description": "alcoholquantity", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "hoursofsleep", 
"description": "hoursofsleep", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "psychologicalstress", 
"description": "psychologicalstress", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "happiness", 
"description": "happiness", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "tiredness", 
"description": "tiredness", 
"type": 1, 
"weight": 1.0 
}, 
{ 
"id": "hormonecycle", 
"description": "hormonecycle", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "fever", 
"description": "fever", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "digestiveillness", 
"description": "digestiveillness", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "medication", 
"description": "medication", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "ambienttemperature", 
"description": "ambienttemperature", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "icrreuse", 
"description": "icrreuse", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "isfreuse", 
"description": "isfreuse", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "bolusreuse", 
"description": "bolusreuse", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "reusedcases", 
"description": "reusedcases", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "typeofbolus", 
"description": "typeofbolus", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "icrsafety", 
"description": "icrsafety", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "isfsafety", 
"description": "isfsafety", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "bolussafety", 
"description": "bolussafety", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "bolususer", 
"description": "bolususer", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "isfuser", 
"description": "isfuser", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "icruser", 
"description": "icruser", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "postprandialminglucose", 
"description": "postprandialminglucose", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "bolusrevise", 
"description": "bolusrevise", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "icrrevise", 
"description": "icrrevise", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "isfrevise", 
"description": "isfrevise", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "expertvalidation", 
"description": "expertvalidation", 
"type": -1, 
"weight": 1.0 
}, 
{ 
"id": "rejectcase", 
"description": "rejectcase", 
"type": -1, 
"weight": 1.0 
}, 
]