A Machine Learning Approach for Specification of Spinal Cord Injuries Using Fractional Anisotropy Values Obtained from Diffusion Tensor Images

Table 2

Results of 𝑡-tests for each feature in the target dataset.

	Feature	𝑝 value = 𝑃  (𝑇≤𝑡)	AVG (patient)	AVG (normal)
1	𝑃	7.69∗10^−9	0.427	0.535
2	𝐴	8.05∗10^−31	0.451	0.684
3	𝐿	1.89∗10^−20	0.475	0.645
4	𝑅	2.39∗10^−26	0.419	0.610
5	(𝑃,𝐴)	3.03∗10^−23	0.439	0.609
6	(𝑃,𝐿)	8.37∗10^−18	0.451	0.590
7	(𝑃,𝑅)	5.61∗10^−19	0.423	0.573
8	(𝐴,𝐿)	8.99∗10^−31	0.463	0.665
9	(𝐴,𝑅)	2.74∗10^−33	0.435	0.647
10	(𝐿,𝑅)	4.35∗10^−31	0.447	0.628
11	(𝑃,𝐴,𝐿)	4.95∗10^−26	0.451	0.621
12	(𝐴,𝐿,𝑅)	2.05∗10^−34	0.448	0.647
13	(𝑃,𝐴,𝑅)	8.57∗10^−27	0.432	0.610
14	(𝑃,𝐿,𝑅)	3.3∗10^−24	0.440	0.597
15	(𝑃,𝐴,𝐿,𝑅)	4.62∗10^−29	0.443	0.619

H₀: the FA value of a normal slice is the same as the FA value of a patient slice; AVG: average of FA value.