Machine learning, medical diagnosis, and biomedical engineering research - commentary

Table 2 Results of training a classifier with SVM (quadratic kernel), using a training set of 400 ultrasound images (half from healthy individuals, half from individuals with Hashimoto’s disease) with a validation set of 100 separate individuals, half of whom had the disease

Attribute number from 1 to 10										*SEN*	*SPC*	*ACC*
(1 - occurs, 0 – does not occur)										*SEN*	*SPC*	*ACC*
1	2	3	4	5	6	7	8	9	10
1	0	1	0	1	0	0	0	0	0	0.831	0.792	0.811
1	1	1	0	1	1	0	1	0	0	0.792	0.831	0.811
1	1	1	0	1	0	1	1	0	0	0.772	0.851	0.811
1	1	1	0	1	0	0	0	0	0	0.831	0.782	0.806
0	1	0	1	1	0	0	0	0	0	0.772	0.841	0.806
1	1	0	1	1	0	0	0	0	0	0.782	0.831	0.806
…
1	1	1	1	1	1	1	1	1	1	0.732	0.811	0.772
…

SEN, SPC, and ACC refer to the sensitivity, specificity, and accuracy of the classifier with the validation set. The accuracy of the classifier does not increase significantly when more attributes are added, implying that some of the attributes contribute little to the performance of the classifier. The table shows the performance of the classifier when constructed from combinations of attributes (attributes used are indicated by 1, not used by 0).

ISSN: 1475-925X