Diagram of Leave One Out Cross-Validation (LOOCV) analysis from which the area under the receiver operating characteristic curve (AUC) and standard error of the AUC were computed from Table I and from which the best candidate scoring function for the Ototoxicity Risk Assessment (ORA) was determined.
Graph of median and inter-quartile range for the indexed (Baseline ) as a function of DPOAE fine structure steps ordered from highest to lowest frequency. The dashed line represents the mean for those subjects with an ASHA-significant change in hearing (inter-quartile range indicated with horizontal shading lines). The solid line represents the mean for those subjects with no change in hearing (inter-quartile range indicated with solid gray shading). The vertical reference line separates the upper and lower quarter octave.
Cross-validated AUC as a function of each candidate scoring functions listed in Table I. The dashed line represents AUC-SE of the most accurate model. ‘a’ indicates the most accurate model and ‘b’ indicates the preferred model according to the ‘One standard error rule’. The quarter-octave model is shown in black triangles (inter-quartile range) and half octaves are shown in gray squares (±1 SE).
Cross-validated ROC curve for the PLS components -Hearing model. The true positive and false positive rates of the ‘6 dB method’ relying on the half-octave (‘◻’) and quarter-octave (‘△’) DPOAE fine structures are also shown. Since the 6 dB method (half- or quarter-octave) is binary and an ROC curve cannot be constructed with these results, they are represented as single points on the figure.
weights used in the final ORA as a function of the highest 12 frequencies measured in octave steps. Positive weights represent a reduction in the DPOAE level re: baseline DPOAE levels while negative weights represent an increase in emission level.
Histogram of the final ORA risk scores for PVs in the study sample. Top panel shows risk scores for PVs without an ASHA-significant hearing change while bottom panel shows PVs with an ASHA-significant hearing change.
Cumulative distribution functions (%) for smoothed measured over 7 visits from 2 control subjects. Gray line represents level changes over highest half octave while black line includes level changes for highest quarter octave.
Candidate scoring functions and number of metrics contained within each function compared in this analysis. Each scoring function was evaluated using both the highest half-octave and highest quarter-octave DPOAE fine structure measurements in 1/48-octave steps.
Study sample patient and treatment characteristics.
Final ORA weights based on partial least-squares (PLS) components -Hearing model. C1, C2, and C3 denote PLS components 1–3.
Article metrics loading...
Full text loading...