Oncology for Dummies!
Weston Carloss
August 26, 2017
Background
Purpose: Provide a simple app to predict survival time for patients with malignant melanoma
Training Data Source: “Melanoma” dataset from R MASS package. Models survival time (in days) for 205 melatoma patients in Denmark from 1962 - 1977 using the following predictors:
- (Anticipated) Patient Health Outcome
- Patient Gender
- Patient Age at Time of Diagnosis
- Year of Tumor Surgery
- Tumor Thickness (Breslow's Thickness)
- Presence/Absence of Tumor Ulceration
App Benefits & Limitations
App Benefits
- Easy to use
- Small, lightweight
- Offers insight into potential benefits of machine learning in cancer outcome prediction
- Built-in documentation
Limitations
- Training data limitations
- Small sample size
- Based on a single country
- Small sample size
- Greatly oversimplifies cancer prediction
Simple Multi-Tab GUI
Model Summary
Call:
lm(formula = .outcome ~ ., data = dat)
Residuals:
Min 1Q Median 3Q Max
-3467.1 -215.1 -69.8 324.5 1973.1
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2162.01 55.94 38.648 < 2e-16 ***
status.Alive 822.04 71.01 11.576 < 2e-16 ***
status.Died_Other_Causes 75.01 63.02 1.190 0.23613
sex.Male -59.72 59.74 -1.000 0.31929
age -5.55 63.44 -0.087 0.93043
year.1965 88.30 74.24 1.189 0.23643
year.1967 45.38 79.72 0.569 0.57021
year.1968 -118.64 83.77 -1.416 0.15907
year.1969 -114.22 84.13 -1.358 0.17692
year.1970 -190.10 84.79 -2.242 0.02665 *
year.1971 -314.56 97.14 -3.238 0.00153 **
year.1972 -512.55 98.03 -5.228 6.64e-07 ***
year.1973 -675.38 93.24 -7.244 3.43e-11 ***
thickness -62.47 67.59 -0.924 0.35707
ulcer.Presence -61.72 67.67 -0.912 0.36348
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 673.6 on 130 degrees of freedom
Multiple R-squared: 0.684, Adjusted R-squared: 0.65
F-statistic: 20.1 on 14 and 130 DF, p-value: < 2.2e-16