TADPOLE

TadpoleCode Repository with the R code in RMD documents used to create the TADPOLE predictions. The repository has all the data as well as the excel file use to generate the data tables

Workflow

Instructions

Processing Workflow:

Data:

1. PredictorsList.txt

            Description: Tab delimeted file with all the features used for prediction
   

2. AllSubjectsWithBasicInfo.txt

            Description: Tab delimeted file with the basic demographics of all TADPOLE subjects
   

3. CompleMRIObs.txt

            Description: Tab delimeted file with the basic longitudinal information of MRI structural information of D1-D2 datasets
            The file was generated in Microsoft Excel
                            - All volumes were transformed by the cubic root
                            - All surface areas were transformed by the square root
                            - Derived columns with the thickness coeficient of variation: StandardDeviation/Thickeness
                            - Derived columns with the strucutral compactness: (square root of Surface Area)/(cubic root of Volume)
                            - Derived columns containing the mean of left and rigth features
                            - Derived columns containing the absolute difference between left and rigth features
                            - Derived columns containing mean thickness and standard deviations of thickness
                            - Derived columns containing mean surface area and standard deviations of transformed areas
                            - Derived columns containing mean volumens and standard deviations of transformed volumens
                            - Derived columns containing the maximum absolute differences between left and right
   

4. D3Clean.txt Description: Tab delimeted file with the basic longitudinal information of MRI structural information of D3 dataset The file was generated in Microsoft Excel - All volumes were transformed by the cubic root - All surface areas were transformed by the square root - Derived columns with the thickness coeficient of variation: StandardDeviation/Thickeness - Derived columns with the strucutral compactness: (square root of Surface Area)/(cubic root of Volume) - Derived columns containing the mean of left and rigth features - Derived columns containing the absolute difference between left and rigth features - Derived columns containing mean thickness and standard deviations of thickness - Derived columns containing mean surface area and standard deviations of transformed areas - Derived columns containing mean volumens and standard deviations of transformed volumens - Derived columns containing the maximum absolute differences between left and right

Processing steps:

1. Data conditioning: Script: TADPOLE_Data_Conditioning.Rmd Description: It will read the CompleteMRIObs.txt and will adjust all features for AGE and ICV it will read the D3Clean.txt and it will impute all missing values the adjsted files then are normalized by the rank inverse transform. Outputs: "TEST_TADPOLEMRI.RDATA" "D3Imputed.csv" "D3Imputed.RDATA" "TRAIN_TADPOLEMRI.RDATA" "TEST_TADPOLEMRImputed.csv" "trainTadploe.norm.RDATA" "testTadploe.norm.RDATA" "D3.norm.RDATA"

2. Predicting conversion: Script: TADPOLE_MCI_to_AD_v4.Rmd Script: TADPOLE_MCI_to_NL_v1.Rmd Script: TADPOLE_NL_to_MCI_v4.Rmd Script: TADPOLE_NL_to_AD_v4.Rmd

            Required files:
                            "PredictorsList.txt"
                            "D3Imputed.RDATA"
                            "TEST_TADPOLEMRI.RDATA"
                            "TRAIN_TADPOLEMRI.RDATA"
                            "trainTadploe.norm.RDATA"      
                            "testTadploe.norm.RDATA"
                            "D3.norm.RDATA"
                           
            Outputs files:
                            *csv files with conversion prediction
   

3. Predicting Time to Event conversion: Script: TADPOLE_Time_to_AD_V2.Rmd Script: TADPOLE_Time_to_MCI_V2.Rmd

            Required files:
                            "PredictorsList.txt"
                            "D3Imputed.RDATA"
                            "TEST_TADPOLEMRI.RDATA"
                            "TRAIN_TADPOLEMRI.RDATA"
                            "trainTadploe.norm.RDATA"      
                            "testTadploe.norm.RDATA"
                            "D3.norm.RDATA"
   
            Outputs files:
                            *csv files with probability of conversion by month
   

4. Ensembling the predictions: Excel file loaded all the prediction of conversion and the estimated time to conversion There the predicted conversion was the multiplication of the predicted conversion probabiliyt by the estimated time of conversion

5. Predicting ADAS conversion:

            Script: TADPOLE_Time_ADAS13_.Rmd
            Script: TADPOLE_Time_ADAS13__MCI.Rmd
            Script: TADPOLE_Time_ADAS13__NL.Rmd
           
            Required files:
                            "PredictorsList.txt"
                            "D3Imputed.RDATA"
                            "TEST_TADPOLEMRI.RDATA"
                            "TRAIN_TADPOLEMRI.RDATA"
                            "trainTadploe.norm.RDATA"      
                            "testTadploe.norm.RDATA"
                            "D3.norm.RDATA"
   
            Outputs files:
                            *csv files with ADAS stratified prediction
   

6. Predicting Ventricle volume: Script: TADPOLE_Time_Ventricle_ICV.Rmd Script: TADPOLE_Time_Ventricle_ICV_MCI.Rmd Script: TADPOLE_Time_Ventricle_ICV_NL.Rmd

            Required files:
                            "PredictorsList.txt"
                            "D3Imputed.RDATA"
                            "TEST_TADPOLEMRI.RDATA"
                            "TRAIN_TADPOLEMRI.RDATA"
                            "trainTadploe.norm.RDATA"      
                            "testTadploe.norm.RDATA"
                            "D3.norm.RDATA"
   
            Outputs files:
                            *csv files with Ventricle stratified prediction