4. Prostate Cancer
One in every seven men will be diagnosed
with Prostate Cancer
Every 6 out of ten men above the age of 65
are diagnosed with proste cancer
One out of every 1 in 38 men will die from
prostate cancer
About 27,540 die annually from prostate
cancer
5. Current Detection
Methods
Digital Rectal Exam
Doctor inserts finger and checks for bumps
or hard areas on prostate
Prostate Specific Antigen blood test
Small amount can be found in blood
if PSA level is above 4 ng/ml, the chance
of prostate cancer goes up.
6. Previous Algorithm
Zlotta:
ANN - Classify when prostate
cancer antigen is less than 10g
Sample Size was 200
The classification accuracy of
the ANN was 92.7%.
8. Methods
The size of dataset was 97
10 different attributes/features
K = 20;
First 75 patients were used for training, the other 22 were
used for testing (for both ANN and KNN).
Labels:
ANN - T,F - True (if patient had cancer) otherwise
False
KNN - 1,2 - 1 = True, 2 = False
11. ANN Conclusion
To ensure error, the ANN was ran 10
times for both the Training and the
Testing Data.
For the Training Data the percent
classification was 99.5%
For the Testing Data the percent
classification was 81.04%
14. KNN conclusions
In conclusion from both figures, the
optimal k value for this dataset was at
k = 1, 19 or 20.
Each value provided a percent
classification of approximately 77%.
15. Conclusion
ANN is a better choice to use for this
dataset because of its high %
classification.
All features could be used for
classification in a clinical setting.
However, further studies should be
done using my code with a larger
dataset to ensure accuracy.