Derived ECG Intervals

1library(yamaa)23adeg <- yamaa_domain("spec.yaml")$output4adeg

1import yamaa23adeg = yamaa.yamaa_domain("spec.yaml").output4adeg

1# Reference solution for the yamaa benchmark adam-adeg-derived-intervals (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)9library(tidyr)1011adeg <- read_csv(12  "/app/input/adeg.csv",13  col_types = cols(AVAL = col_double(), .default = col_character())14)1516wide <- adeg |>17  select(STUDYID, USUBJID, AVISIT, PARAMCD, AVAL) |>18  pivot_wider(names_from = PARAMCD, values_from = AVAL)1920# A QTc record needs both QT and RR; RR in seconds is RR in ms over 1000.21qtc <- wide |>22  filter(!is.na(QT), !is.na(RR)) |>23  mutate(24    RRSEC = RR / 1000,25    QTCBR = QT / sqrt(RRSEC),26    QTCFR = QT / (RRSEC^(1 / 3))27  )2829qtcbr <- qtc |>30  transmute(31    STUDYID, USUBJID, AVISIT,32    PARAMCD = "QTCBR",33    PARAM = "QTcB - Bazett's Correction Formula Rederived (ms)",34    AVAL = QTCBR,35    AVALU = "ms"36  )3738qtcfr <- qtc |>39  transmute(40    STUDYID, USUBJID, AVISIT,41    PARAMCD = "QTCFR",42    PARAM = "QTcF - Fridericia's Correction Formula Rederived (ms)",43    AVAL = QTCFR,44    AVALU = "ms"45  )4647# An RRR record needs a present, nonzero heart rate.48rrr <- wide |>49  filter(!is.na(HR), HR != 0) |>50  transmute(51    STUDYID, USUBJID, AVISIT,52    PARAMCD = "RRR",53    PARAM = "RR Duration Rederived (ms)",54    AVAL = 60000 / HR,55    AVALU = "ms"56  )5758adeg_out <- bind_rows(59  adeg |> select(STUDYID, USUBJID, PARAMCD, PARAM, AVISIT, AVAL, AVALU),60  qtcbr,61  qtcfr,62  rrr63) |>64  select(STUDYID, USUBJID, PARAMCD, PARAM, AVISIT, AVAL, AVALU)6566dir.create("/app/output", showWarnings = FALSE)67write_csv(adeg_out, "/app/output/adeg.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adeg-derived-intervals (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67import math8from pathlib import Path910import polars as pl111213def format_number(value: float | None) -> str | None:14    if value is None or not math.isfinite(value):15        return None16    if float(value).is_integer():17        return str(int(value))18    return repr(float(value))192021base = pl.read_csv("/app/input/adeg.csv", infer_schema=False)22numeric = base.with_columns(AVALNUM=pl.col("AVAL").cast(pl.Float64, strict=False))2324wide = numeric.group_by(["STUDYID", "USUBJID", "AVISIT"]).agg(25    QT=pl.col("AVALNUM").filter(pl.col("PARAMCD") == "QT").first(),26    RR=pl.col("AVALNUM").filter(pl.col("PARAMCD") == "RR").first(),27    HR=pl.col("AVALNUM").filter(pl.col("PARAMCD") == "HR").first(),28)2930collected = base.select(31    "STUDYID", "USUBJID", "PARAMCD", "PARAM", "AVISIT", "AVAL", "AVALU"32)3334new_rows: list[dict] = []35for record in wide.to_dicts():36    studyid, usubjid, avisit = record["STUDYID"], record["USUBJID"], record["AVISIT"]37    qt, rr, hr = record["QT"], record["RR"], record["HR"]38    # A QTc record needs both QT and RR; RR in seconds is RR in ms over 1000.39    if qt is not None and rr is not None:40        rr_sec = rr / 100041        new_rows.append(42            {43                "STUDYID": studyid,44                "USUBJID": usubjid,45                "PARAMCD": "QTCBR",46                "PARAM": "QTcB - Bazett's Correction Formula Rederived (ms)",47                "AVISIT": avisit,48                "AVAL": format_number(qt / math.sqrt(rr_sec)),49                "AVALU": "ms",50            }51        )52    if qt is not None and rr is not None:53        rr_sec = rr / 100054        new_rows.append(55            {56                "STUDYID": studyid,57                "USUBJID": usubjid,58                "PARAMCD": "QTCFR",59                "PARAM": "QTcF - Fridericia's Correction Formula Rederived (ms)",60                "AVISIT": avisit,61                "AVAL": format_number(qt / (rr_sec ** (1 / 3))),62                "AVALU": "ms",63            }64        )65    # An RRR record needs a present, nonzero heart rate.66    if hr is not None and hr != 0:67        new_rows.append(68            {69                "STUDYID": studyid,70                "USUBJID": usubjid,71                "PARAMCD": "RRR",72                "PARAM": "RR Duration Rederived (ms)",73                "AVISIT": avisit,74                "AVAL": format_number(60000 / hr),75                "AVALU": "ms",76            }77        )7879# Collected rows first, then QTCBR, QTCFR, and RRR per visit.80order = {"QTCBR": 0, "QTCFR": 1, "RRR": 2}81new_rows.sort(82    key=lambda r: (order[r["PARAMCD"]], r["USUBJID"], r["AVISIT"])83)84derived = pl.DataFrame(85    new_rows,86    schema={87        "STUDYID": pl.String,88        "USUBJID": pl.String,89        "PARAMCD": pl.String,90        "PARAM": pl.String,91        "AVISIT": pl.String,92        "AVAL": pl.String,93        "AVALU": pl.String,94    },95)9697adeg = pl.concat([collected, derived], how="diagonal").select(98    "STUDYID", "USUBJID", "PARAMCD", "PARAM", "AVISIT", "AVAL", "AVALU"99)100101Path("/app/output").mkdir(exist_ok=True)102adeg.write_csv("/app/output/adeg.csv")

1/* opensas reference solution for the yamaa benchmark adam-adeg-derived-intervals.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data adeg;6  length STUDYID USUBJID PARAMCD PARAM AVISIT AVALU $1024;7  infile "/app/input/adeg.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. PARAMCD : $char1024. PARAM : $char1024. AVISIT : $char1024. AVAL AVALU : $char1024.;9run;1011proc sql;12  create table qt as select a.*,b.AVAL as RRVAL from adeg as a inner join adeg as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.AVISIT=b.AVISIT where a.PARAMCD='QT' and b.PARAMCD='RR' and not missing(a.AVAL) and not missing(b.AVAL);13quit;14data corrected;15  length PARAMCD PARAM AVALU $1024;16  set qt;17  QTVAL=AVAL; AVALU='ms';18  PARAMCD='QTCBR'; PARAM="QTcB - Bazett's Correction Formula Rederived (ms)"; AVAL=QTVAL/sqrt(RRVAL/1000); output;19  PARAMCD='QTCFR'; PARAM="QTcF - Fridericia's Correction Formula Rederived (ms)"; AVAL=QTVAL/(RRVAL/1000)**(1/3); output;20run;21data heart;22  length PARAMCD PARAM $1024;23  set adeg;24  if PARAMCD='HR' and not missing(AVAL) and AVAL ne 0;25  AVAL=60000/AVAL; PARAMCD='RRR'; PARAM='RR Duration Rederived (ms)'; AVALU='ms';26run;27data result; set adeg corrected heart; run;2829proc sql;30  create table final as select STUDYID, USUBJID, PARAMCD, PARAM, AVISIT, AVAL, AVALU from result;31quit;32proc export data=final outfile="/app/output/adeg.csv" dbms=csv replace; run;