Survival Boosting Model for Hep-C
Overview
Input and Output
KO input
The input to the KO is specified in the KO’s service specification file service.yaml
.
The following shows an example input:
{
"GENDERM": true,
"RACEHISPANIC": false,
"RACEMISSING": false,
"RACEOTHER": false,
"RACEWHITE": false,
"AGE": 52,
"ALB_MAX": 4.2,
"ALB_MIN": 3.7,
"ALB_DIFFMAX": 0.07247,
"ALB_DIFFMIN": -0.0801,
"ALB_TVAR": 0.07657,
"ALB_LAST": 4,
"ALKRATIO_MAX": 0.7477,
"ALKRATIO_MIN": 0.5604,
"ALKRATIO_DIFFMAX": 0.0252,
"ALKRATIO_DIFFMIN": -0.0297,
"ALKRATIO_TVAR": 0.02729,
"ALKRATIO_LAST": 0.6429,
"ALTRATIO_MAX": 1.2444,
"ALTRATIO_MIN": 0.6508,
"ALTRATIO_DIFFMAX": 0.0765,
"ALTRATIO_DIFFMIN": -0.0947,
"ALTRATIO_TVAR": 0.08713,
"ALTRATIO_LAST": 0.8571,
"ASTRATIO_MAX": 1.3409,
"ASTRATIO_MIN": 0.75,
"ASTRATIO_DIFFMAX": 0.079,
"ASTRATIO_DIFFMIN": -0.0949,
"ASTRATIO_TVAR": 0.08856,
"ASTRATIO_LAST": 0.9512,
"BIL_MAX": 0.8,
"BIL_MIN": 0.4,
"BIL_DIFFMAX": 0.05743,
"BIL_DIFFMIN": -0.0636,
"BIL_TVAR": 0.5938,
"BIL_LAST": 0.6,
"BUN_MAX": 17,
"BUN_MIN": 10,
"BUN_DIFFMAX": 1.4194,
"BUN_DIFFMIN": -1.368,
"BUN_TVAR": 1.3135,
"BUN_LAST": 13,
"CI_MAX": 106,
"CI_MIN": 101,
"CI_DIFFMAX": 1.0871,
"CI_DIFFMIN": -1.0231,
"CI_TVAR": 0.9979,
"CI_LAST": 103,
"CRE_MAX": 1.1,
"CRE_MIN": 0.9,
"CRE_DIFFMAX": 0.04443,
"CRE_DIFFMIN": -0.04348,
"CRE_TVAR": 0.0398,
"CRE_LAST": 1,
"GLU_MAX": 12,
"GLU_MIN": 8.7,
"GLU_DIFFMAX": 0.8696,
"GLU_DIFFMIN": -0.949,
"GLU_TVAR": 0.8717,
"GLU_LAST": 10.1,
"HEM_MAX": 15.6,
"HEM_MIN": 14,
"HEM_DIFFMAX": 0.2306,
"HEM_DIFFMIN": -0.291,
"HEM_TVAR": 0.2682,
"HEM_LAST": 14.8,
"PLT_MAX": 25.7,
"PLT_MIN": 19.5,
"PLT_DIFFMAX": 0.9277,
"PLT_DIFFMIN": -1.01,
"PLT_TVAR": 1.01,
"PLT_LAST": 22.2,
"K_MAX": 4.6,
"K_MIN": 3.9,
"K_DIFFMAX": 0.1505,
"K_DIFFMIN": -0.1513,
"K_TVAR": 0.143,
"K_LAST": 4.2,
"NA_MAX": 141,
"NA_MIN": 137,
"NA_DIFFMAX": 0.9318,
"NA_DIFFMIN": -0.9587,
"NA_TVAR": 0.8822,
"NA_LAST": 139,
"TOTP_MAX": 7.9,
"TOTP_MIN": 7.1,
"TOTP_DIFFMAX": 0.1098,
"TOTP_DIFFMIN": -0.1295,
"TOTP_TVAR": 0.1243,
"TOTP_LAST": 7.5,
"WBC_MAX": 8.4,
"WBC_MIN": 5.8,
"WBC_DIFFMAX": 0.4184,
"WBC_DIFFMIN": -0.4348,
"WBC_TVAR": 0.4439,
"WBC_LAST": 6.8,
"APRI_MAX": 0.6334,
"APRI_MIN": 0.3301,
"APRI_DIFFMAX": 0.04769,
"APRI_DIFFMIN": -0.0533,
"APRI_TVAR": 0.059,
"APRI_LAST": 0.4405,
"ASTALT_MAX": 1.0455,
"ASTALT_MIN": 0.7308,
"ASTALT_DIFFMAX": 0.04751,
"ASTALT_DIFFMIN": -0.0473,
"ASTALT_TVAR": 0.0477,
"ASTALT_LAST": 0.88235,
"BMI_MAX": 27.87,
"BMI_MIN": 25.36,
"BMI_DIFFMAX": 0.8355,
"BMI_DIFFMIN": -0.743,
"BMI_TVAR": 0.6057,
"BMI_LAST": 26.67,
"TimeToOutcome": 365,
"Outcome": 0.5
}
KO output
The KO output as the result from the model is the log of the hazard ratio:
{"result": -5.95 }
Activate the KO and try out its service
Please refer to the documentation of Rplumber-Server
TO DO list
The current work creates a prototype Knowledge Object based on the Survival Boosting Model for Hep-C prediction.
More work will be needed for bringing the KO closer to the real-world practice, including but definitely not limited by the following:
- Understand the raw data and the workflow to request and aggregate the data;
- Understand the data format and context so that proper summary statistics can be computed for each variable;
- Implement the procedural code to process the raw data and generate the input in the specified format;
- Implement the presentation/transformation/interpretation of the model output
- Enhance the documentation of the KO(s) and the deployment instruction of the KO(s)