Evaluation of Deformable Image Registration Methods for Dose Monitoring in Head and Neck Radiotherapy

Table 3

Average cumulated dose (Gy) error by landmarks and by registration methods with the first observer as anatomical reference.

	Landmark dose (Gy)
	Bones							Soft tissues							Precision (AVG)	Accuracy (AVG SD)
	1	2	3	4	5	6	7	8	9	10	11	12	13	14
Planned dose	44.47	35.13	54.87	64.91	63.64	8.84	44.50	64.74	21.60	66.87	66.91	63.95	39.71	41.37	48.68	10.35
Cumulated dose difference^*	1.95	1.96	1.68	1.96	1.71	3.09	2.87	2.55	2.52	1.93	0.69	1.12	5.17	4.25	2.39	2.40
	Landmark Cumulated Dose Error (Gy)
Interobserver variability	0.41	0.23	0.29	0.38	0.67	0.26	0.28	0.31	0.41	0.77	0.14	0.61	1.68	3.14	0.68	0.75
FFD MI filtered CTs	0.36	0.60	0.43	0.58	0.69	0.43	0.41	0.59	0.63	0.74	0.91	0.48	2.63	2.41	0.85	0.93
Demons MI filtered CTs	0.40	0.39	0.60	0.57	0.71	0.33	0.45	0.41	1.09	0.75	0.69	0.78	2.74	2.46	0.88	0.95
Demons MI	0.33	0.39	0.57	0.54	0.69	0.38	0.48	0.47	1.11	0.90	0.73	0.80	2.56	2.33	0.88	0.92
Demons MSE	0.41	0.52	0.42	0.48	0.88	0.40	0.43	0.56	0.85	1.38	0.74	0.75	3.15	2.96	0.99	1.22
FFD MSE filtered CTs	0.49	0.69	0.52	0.59	0.86	0.55	0.38	0.83	0.95	0.83	0.98	0.57	2.37	3.09	0.98	1.09
FFD MI	0.31	0.63	0.47	0.53	0.53	0.48	0.61	0.82	1.09	0.84	0.85	0.85	2.45	2.68	0.94	1.02
Demons MSE filtered CTs	0.41	0.52	0.42	0.48	0.88	0.40	0.43	0.56	0.85	1.38	0.74	0.75	3.15	2.96	0.99	1.25
FFD MSE	0.34	1.16	1.13	1.27	1.02	0.45	0.92	1.34	1.88	1.26	1.06	1.86	2.75	3.91	1.45	1.60
Demons D. maps	0.99	1.29	3.11	1.19	1.24	0.53	0.94	1.28	1.76	2.39	1.29	2.22	3.05	5.14	1.89	2.14
FFD D. maps	0.95	1.56	1.24	0.94	1.53	0.53	1.00	1.54	2.66	2.90	1.32	2.54	2.99	4.87	1.90	2.18
Rigid MSE	0.94	1.67	1.20	2.78	1.80	0.59	1.02	1.71	3.35	3.99	1.35	2.61	3.44	4.27	2.19	2.60

AVG: average, AVG SD: average standard deviation, FFD: free form deformation, MSE: mean squared error, MI: mutual information, and CT: computed tomography (scan).
Methods are classified by their performance order. The performance is defined by the accuracy (average of the cumulated dose error) and the precision (average of the standard deviation cumulated dose error). The cumulated dose difference (*) represents the reference difference between planned and cumulated doses from the first observer. A second observer allows quantifying the interobserver variability.
The “FFD with MI on filtered CTs” errors are inferior to all the methods errors ( < 0.03), except for the “demons with MI on filtered CTs” and “demons with MI on original CTs” methods and for the “demons with MSE on original CTs” ( = 0.06). The “demons with MI on filtered CTs” errors are inferior to the “delineation maps based method” errors, the “demons with MSE on filtered CTs” errors, and the “FFD MSE on original CTs” errors (respectively, < 0.01, ≤ 0.03 and < 0.01).