Operation of a P300-based brain-computer interface in patients with Duchenne muscular dystrophy

Abstract

A brain-computer interface (BCI) or brain-machine interface is a technology that enables the control of a computer and other external devices using signals from the brain. This technology has been tested in paralysed patients, such as those with cervical spinal cord injuries or amyotrophic lateral sclerosis, but it has not been tested systematically in Duchenne muscular dystrophy (DMD), which is a severe type of muscular dystrophy due to the loss of dystrophin and is often accompanied by progressive muscle weakness and wasting. Here, we investigated the efficacy of a P300-based BCI for patients with DMD. Eight bedridden patients with DMD and eight age- and gender-matched able-bodied controls were instructed to input hiragana characters. We used a region-based, two-step P300-based BCI with green/blue flicker stimuli. EEG data were recorded, and a linear discriminant analysis distinguished the target from other non-targets. The mean online accuracy of inputted characters (accuracy for the two-step procedure) was 71.6% for patients with DMD and 80.6% for controls, with no significant difference between the patients and controls. The P300-based BCI was operated successfully by individuals with DMD in an advanced stage and these findings suggest that this technology may be beneficial for patients with this disease.

Figure 1

A region-based two-step P300-based hiragana speller. The subject was required to count the number of intensifications of the green/blue circle containing the target character to be input. Each circle flashed eight times, and the interval between two flashes was 175 ms, consisting of 100 ms of intensification (green) and 75 ms of rest (blue). We divided the 6 × 9 matrix into six circled regions including nine characters each. First, each region was intensified individually. When a region that included a target character was selected, the speller matrix moved to the second step. The second step used 3 × 3 regions with one character each, and each region was again intensified.

Figure 2

Online and offline accuracy in the BCI operation. (a) Online accuracy in the region-based two-step P300-BCI operation, which indicated the accuracy of both steps in total, was 79.8% for patients with DMD and 83.4% for controls; no significant difference was observed between them (t-test, p = 0.69). Each triangle (DMD) and circle (controls) indicates each data point. (b) Offline accuracy for both groups in each flashing sequence was plotted (triangle: DMD, and circle: controls).

Figure 3

Performance on the character input. (a) Online accuracy of inputted characters (accuracy for the two-step procedure) in the DMD patients was 71.6% and that in the control was 80.6%, with no significant difference between them (t-test, p = 0.32). Each triangle (DMD) and circle (controls) indicates a data point. (b) Offline accuracy for both groups in each flashing sequence was plotted (triangle: DMD, and circle: controls).

Figure 4

EEG waveforms in DMD and controls. EEG waveforms obtained from P4 electrodes, averaged for patients with DMD and controls, are shown. The 700 ms waveforms extracted from the onset of intensification were averaged. Solid lines indicate the target ERPs and dashed lines indicate non-target ERPs. In the controls, two positive peaks are shown at approximately 243 and 535 ms after the onset of stimuli, whereas in the DMD group, no clear peak corresponding to the early component was evident, but a blurred peak was seen at approximately 450–550 ms.