Intraoperative x‐ray to measure distance between DBS leads: A reliability study

Many factors can jeopardize the accuracy of deep brain stimulation (DBS) lead placement. Confirmation of lead placement while the patient is still in the operating room would be advantageous. Intraoperative MRI or CT can identify placement errors, but these modalities can be cost‐ or time prohibitive. Intraoperative fluoroscopy may give information on the accuracy of the Y coordinate, but the accuracy of the X coordinate usually cannot be confirmed. When an object of known dimensions is present in the brain, such as a unilateral DBS lead, its dimensions can be used to calculate unknown distances. The objective of this study was to determine if intraoperative AP skull x‐ray accurately predicts the distance between DBS electrodes using postoperative MRI as the gold standard.


A B S T R A C T
Background: Parkin mutations in patients with earlyonset Parkinson's disease (EOPD) are estimated to occur in 49% of familial cases and 18% of sporadic cases. Methods: We analyzed the entire sequence-coding region and dosage mutations of parkin in 63 Mexicanmestizo EOPD patients and 120 controls. Results: Parkin mutations were present in 34 patients (54.0%). Exon rearrangements, predominantly spanning exons 9 and 12 (31.7% and 19.0%, respectively) were present in 32 patients, with 17.5% carrying simple heterozygous and 25.4% carrying compound heterozygous parkin mutations.
Conclusions: A higher frequency of parkin exon rearrangements than of sequence mutations was observed. Patients with parkin exons 9 and 12 rearrangements showed a later age at onset than did cases with other regions affected (40.3 6 4.5 vs 30.1 6 8.8; P 5 .005), suggesting a mutational hot spot in the etiology of Mexican-mestizo patients with EOPD. To our knowledge, this study represents the largest sampling of Mexican-mestizo patients with EOPD cases for which parkin sequence and dosage alterations were analyzed. Parkinson's disease (PD) is characterized by resting tremor, bradykinesia, rigidity, and postural instability. It affects approximately 1% of the population > 65 years old: when present in those < 45 years old, it is referred to as early-onset PD (EOPD). 1 Most PD cases are sporadic, but approximately 5%-10% are the result of Mendelian inherited mutations. Mutations in parkin (PARK2; MIM#600116) are the most important genetic cause of EOPD with autosomal recessive inheritance. 2 The PARK2 mutation frequency in EOPD is approximately 49% in familial cases and about 18% in cases without family history of PD. 3 To date, 247 different PARK2 mutations, comprising simple mutations and copy number variations (CNVs) of single or multiple exons, have been reported. 4 PARK2 spans %1.4 Mb and is on 6q25.2-q27. 5 PARK2 codes for parkin, a ubiquitin E3 ligase that tags specific substrates for protein degradation through the ubiquitin-proteasome system. 6 PARK2 mutations, either homozygous, compound heterozygous, or heterozygous, have been identified in familial and sporadic patients of different ethnicities. 7 However, the relevance of heterozygous mutations in PARK2 as a risk factor for PD remains controversial. [8][9][10] The role of PARK2 mutations has been studied in PD patients from Europe, 11,12 Asia, [13][14][15] and North America, 16,17 but data for Mexican patients are very scarce. 8,18 The goal of this study was to determine the prevalence and type of PARK2 gene mutations in a sample of Mexican-mestizo patients with EOPD.
National Institute of Neurology and Neurosurgery (NINN) were enrolled in this study. The mean age at onset (AAO) was 37.3 6 6.6 years (range, 16-45 years). Two neurologists with experience in movement disorders established the clinical diagnosis according to the UK Parkinson's Disease Society Brain Bank criteria. 19 Control subjects were 120 healthy, unrelated volunteers (73 females) who had no family history of movement disorders development and, to minimize the possibility of PD development, were ! 45 years (range, 45-82 years). Patients and controls originated from Mexico City and surrounding areas, were Mexicanmestizo for at least 3 generations, and had a similar socioeconomic level. The study was approved by the NINN institutional review board. All par-ticipating individuals signed an informed consent form.
Genomic DNA was isolated from whole blood using standard techniques. The PARK2 gene was sequenced using published primers. 20 The 12 exons of PARK2 were examined in triplicate for dosage abnormalities, as described. 21 Negative controls and a commercial control DNA were included in each run; all positive results were confirmed at least twice. Heterozygosity was inferred, as chromosomal phase was not determined. Statistical analyses were performed using SPSS software v. 15.0 (SPSS, Chicago, IL). Quantitative data were analyzed by the Student t test. Frequency data were compared between groups by the v 2 test or Fisher's exact test. P < .05 was considered statistically significant.  45 14 a n/a n/a WT 38.3 6 6.4 À 46-55 10 a n/a n/a WT 35.0 6 7.2 þ 56-60 5 a n/a n/a WT 37.2 6 6.3 ?

Results
Of the 63 PD patients studied, 45 (71.4%) began the disease with tremor and 18 (28.6%) with rigidity. PARK2 mutations were not associated with tremor or rigidity, or with positive or negative family history (data not shown). In 34 cases (54%) from 31 families, PARK2 mutations were identified and shown to be homozygous in 7 cases representing 4 families, compound heterozygous in 16 (16 families), and heterozygous in 11 (11 families). Exon rearrangements were present in 32 of 34 patients (94.1%), whereas 2 of 34 had a sequence mutation (5.9%). PARK2 exon deletions were more frequent than were exon duplications (93.7% and 6.3%, respectively; Table 1).
In family 1, 2 affected siblings had a homozygous point deletion (c.155delA; Fig. 1A). Both parents and the 6 unaffected siblings were not available for the study. In family 3 (Fig. 1B), 3 affected siblings showed a homozygous deletion of exon 5. The hus-band of the propositus had a sister with PD who was also homozygous for exon 5 deletion; he was heterozygous for the same mutation. Although their families are considered unrelated, the size of the small town ($648 habitants) in which they reside suggests possible endogamy. For genetic counseling of this family, it is very important to consider that the progeny of II.3 and II.4 has a 50% risk of developing PD. To date, none has developed PD symptoms, with the eldest being 27 years old. The pedigrees of families 1 and 3 are compatible with autosomal recessive inheritance.

Discussion
To the best of our knowledge, this study represents the largest sample size of Mexican-mestizo EOPD cases in which the complete sequence and dosage alterations of PARK2 were analyzed. There are only 2 previous studies reporting PARK2 mutations in Mexican PD patients, each with a very small sample. 10,18 Our finding that 54.0% of the patients had PARK2 mutations, with 94.1% dosage mutations and 5.9% point-sequence mutations, evidenced Mexican-mestizo EOPD patients as exhibiting the highest frequency of PARK2 rearrangements reported so far.
The frequency of dosage mutations varies according to the population studied and can account for 33%-67% of all mutations. 3,12,22,23 This high rate of rearrangements is due to the location of the PARK2 gene within the common fragile site, FRA6E. 5 As in other studies, 10,13,15 deletions were more frequent than were duplications (93.7% and 6.3%, respectively). Deletions in exons 3 and 4 have been reported with a high frequency and are considered a mutational hot spot. 10,24 Although we found 17.5% to be exon 4 rearrangements, the most frequent rearrangements were detected in exons 9 and 12 (31.7% and 19.0%, respectively), suggesting a second hot spot in these exons. In a study of PD cases in India, a mutation frequency of 7.2% and 11.5% for deletions in exons 9 and 12, respectively, was reported. 25 Exon 9 deletions were present in 2.3% of PD cases in the South African population. 26 In addition, here, 3 new exon rearrangements (exons 9-10 deletion; exons 9-12 deletion; and exons 3-8 duplication) were identified, thereby adding to the growing list of known PARK2 mutations.
We found a high frequency of sporadic cases with PARK2 mutations (61.1%) compared with 18% and 33.8% reported for whites. 3,23 The sporadic cases had greater risk to have exons 9-12 rearrangements than did familial cases, highlighting the importance to study PARK2 in sporadic EOPD cases.
We identified a homozygous point mutation (c.155delA) in only 1 family with 2 affected members. This mutation has a high prevalence in Spanish 3,27,28 and Hispanic PD patients, 10 suggesting an ancestral European origin. 28 Our value (17.5%) for the frequency of simple heterozygous CNVs was within the range (2.3%-28.0%) for EOPD cases. 12,29 Recent evidence demonstrated that a single CNV (but not a single missense mutation) in the PARK2 locus could be validated molecularly and associated with PD susceptibility. 10,29,30 Our results are compatible with this hypothesis, although in our patients, the possibility of a second, undetected parkin mutation (promoter region) or a mutation in another PD gene cannot be excluded. Low frequencies of heterozygous CNVs have been reported for healthy individuals. [30][31][32][33] Because no data were available for the Mexican population, we screened 120 controls for PARK2 dosage; a very low frequency (0.83%) was observed.
Contrary to reports in which PARK2 mutations showed an earlier AAO compared with cases without mutations, 3,10,12,29 in our sample the mean AAO was similar in cases with and without mutations. Furthermore, we did not observe any statistically significant difference between carriers of 1 or 2 PARK2 mutations with AAO, with positive or negative family history, or with tremor or rigidity. A larger sample size is needed to confirm these observations.
Here, an association between later AAO and dosage mutation in the IBR-RING2 region was observed. Further functional studies are needed to explain these results. It has been suggested that multiple differential processes, induced by diverse parkin mutations depending on the location of the functional domain, may underlie the development of PD. 34 The RING1-IBR-RING2 structure of parkin is critical for intrinsic ubiquitin E3 ligase activity but is not necessary for activation of the 26S proteasome. 34 Studies in 2 animal models showed that lack of IBR-RING2 domains provokes progressive dopaminergic neuron degeneration in transgenic Drosophila 35 and late onset in transgenic mice. 36 The high prevalence of PARK2 dosage mutations observed in our sample indicates a change in the strategy for molecular diagnosis: CNV should be screened before mutations are sequenced, at least in the Mexican population. 25 Cholinergic deficiency occurs in Parkinson's disease (PD). It is severe and extends to frontal cortical areas in PD dementia (PDD). 1,2 In vivo measurement of cortical cholinergic activity is possible by studying short-latency afferent inhibition (SAI). 3 It enables the measurement of the modulation of excitability of the motor cortex by employing a time-locked coupling of peripheral nerve and motor cortex stimulation. 3,4 In this study, we aimed to use SAI to examine differential cholinergic activity in PD patients with or without dementia, in comparison to Alzheimer's disease (AD) patients and controls.

Subjects
Patients with PD, PDD, and AD (10 patients in each group) were consecutively recruited from our outpatient clinic during 2009 to 2011. Controls were spouses or relatives of the patients (n ¼ 10). Diagnosis of AD and PDD was performed according to National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer's Disease and related Disorders Association 5 and Movement Disorder Society criteria for PDD, 6 respectively. Inclusion criteria are summarized in Supporting Table 1. All subjects or relatives gave written consent. The protocol was approved by the Hacettepe University Ethics Committee. Cranial MRI and/or CT were obtained in all of the patients. None of the cases had structural or clinically relevant vascular lesions.
Dementia cases were de novo cases with mild to moderate severity (Mini-Mental State Examination [MMSE] scores of 10-24) and were not receiving any antidementia medication. None of the subjects were taking any cholinergic or anticholinergic drugs during and for at least 2 weeks before the study. PD and PDD cases were free of antiparkinsonian treatment for at least 12 hours before and during the SAI study.

SAI Study
Transcranial magnetic stimulation (TMS) was performed by the MagPro stimulator (Medtronic A/S, Copenhagen, Denmark). A parabolic coil, which can induce focal cortical stimulation as a figure-of-eight coil, 7 with an external diameter of 120 mm, was held over the motor cortex at an optimum position for eliciting motor-evoked potentials (MEPs) in the contralateral first dorsal interosseous (FDI) muscle. The direction of induced current was from posterior to anterior. In PD and PDD cases, the hemisphere contralateral to the more affected body side was stimulated; in the others, the left hemisphere was stimulated. Contralateral median nerve stimuli (MNS) at the wrist with an intensity that produced only a slight thumb twitch were used as conditioning stimuli. The stimulation rate was 3 Hz. A total of 250 responses were averaged to obtain a somatosensory-evoked potential (SEP) signal.
The SAI examiner was blind to the study groups. The following steps were performed to obtain SAI. First, the MEP in the FDI was elicited by TMS of the motor cortex at rest. The electromyography signal was monitored by loudspeakers. The response was called the control MEP (MEP c ). The amplitude of MEP c was measured as peak-to-peak amplitude (Supporting Fig. 1). Second, MNS was performed to obtain SEP. N20 latency was measured. One to eight milliseconds was added to the N20 latency to determine interstimulus intervals (ISIs). Third, MNS was followed by TMS to the contralateral motor area with the predetermined ISIs. The obtained MEP amplitude was expressed as the percentage of MEP c . This was referred as test MEP (MEP t1-t8 ). Three trials were repeated at each ISI; control and test trials were randomized. The grand mean of the test MEPs (MEP tmean ) was the average of MEP t1-t8 .

Neuropsychological Examination
Neuropsychological tests administered are listed in Table 1. The time (in seconds) for completing the task was noted for reciting months backward and trailmaking tests. Scores of the other tests were presented as the total number of correct responses, with higher scores indicating better responses. The administration and scoring of these tests have been described previously. [8][9][10][11][12][13][14] Statistical Analysis Values were reported as mean 6 standard error. SAI and the remaining numerical data were compared by means of one-way analysis of variance (ANOVA). Bonferroni's post-hoc comparisons were conducted to examine differences between the two subject groups, if the ANOVA revealed a significant effect (P < 0.05). The effect of groups on the change in SAI by time was investigated using repeated-measures ANOVA. Greenhouse-Geisser's correction was used when sphericity assumption was violated. An overall 5% type 1 error Error bars indicate standard errors. In the PD group, MEP t values were similar to controls at each ISI (P > 0.05), whereas in the AD and PDD groups, SAI was significantly reduced with higher MEP t values, compared to controls, at each ISI until N2018 ms (for MEP t1-t7 , F (3,36) 5 3.63-10.07; for all F values, P < 0.02; for MEP t8 , F (3,36) 5 1.11 and P 5 0.36). Repeated-measures ANOVA showed that SAI courses were similar in PDD and AD groups and were significantly different from PD patients and controls (F (3) 5 13.13; P < 0.0001). For all groups, SAI responses changed significantly with time (F (5) : 15.49; P < 0.0001). (B) The scatterplot shows individual MEP tmean values of PD, PDD, AD, and control groups. Amplitude of the MEP t was reported as a %MEP c . The line represents the upper limit of the control (mean 1 2SD). As seen from the diagram, all of the PD cases but 1 (90%) had normal MEP t (lower than the upper limit of control). On the other hand, only 3 of the PDD (30%) and 2 of the AD cases (20%) had MEP t values lower than the upper limit of the control. (C) The scatterplot shows MMSE and MEP tmean values of individual cases in all study groups. MMSE and SAI were inversely related (r 5 20.68; P < 0.0001). Patients with lower MMSE had higher MEP tmean values, implying that SAI is more impaired in cases with higher cognitive dysfunction. level was used for statistical significance. A chi-square test was used for comparison of categorical data. Spearman's correlations for rank data were used for correlation analysis of neuropsychological test scores and SAI responses. Correction for multiple analyses was performed by adjusting the significance level to P < 0.004.

Results
Demographic and clinical characteristics of the subjects are presented in Table 1. Age and education were similar in the four groups. Duration of parkinsonism was longer (8.4 6 1.6 versus 2.3 6 0.5; P ¼ 0.004), motor scores were more severe (22.9 6 1.7 versus 9.8 6 1.6; P < 0.001), and dosages of dopaminergic medications were higher (943 6 147.2 versus 345.5 6 77.7; P ¼ 0.003) in PDD than PD subjects. Visual hallucinations (VHs) were present in 4 of the PDD and in 1 of the PD, but in none of the AD patients.
Neuropsychological test performances are summarized in Table 1. MMSE scores in AD and PDD subjects were comparable (18.8 6 1.3 versus 19.8 6 0.7; P > 0.05) and lower than in PD and control cases (P < 0.0001). In all cognitive domains tested, PDD and AD subjects performed worse than PD patients and controls. In the comparison of PDD and AD, episodic memory impairment was more marked in AD (P ¼ 0.03). Geriatric Depression Scale scores were higher in PDD than PD (P < 0.004).
Spearman's correlation analysis of MMSE and MEP tmean disclosed the presence of a negative correlation (r ¼ À0.68; P < 0.0001) (Fig. 1C). A similar relation was also evident in neuropsychological tests measuring attention, executive functions, memory, and visuospatial function (Supporting Table 3).

Discussion and Conclusion
This study demonstrates that SAI is impaired in PD patients with established dementia, but not in cases without dementia. The degree of impairment is as severe as that observed in AD. Furthermore, the degree of impairment is higher in cases with higher cognitive dysfunction. The close association of cognitive functions and SAI is interesting because both were assumed to depend, to some extent, on the intact functioning of cortical cholinergic neuromodulation, albeit in different cortical regions. Of note, these findings may not be generalized to tremor-dominant patients because they were not represented in our study. SAI was introduced as an in vivo tool for evaluating functional cortical cholinergic activity. 3 It is markedly decreased in AD 15,16 and with muscarinic receptor blockage in healthy subjects, 3 but not in disorders without overt cholinergic deficiency. [17][18][19][20] Our results are in compliance with these findings, because marked reduction of SAI was obtained in AD subjects.
Cortical cholinergic deficiency also occurs in PD. 1,2,21 Diminished dopaminergic transmission is the key feature. Therefore, SAI was studied in PD patients in ''off'' and ''on'' conditions. In the ''off'' condition, SAI was normal or increased, 19,22 whereas in the ''on'' condition, it was impaired. 22 In our study, we also replicated these findings by demonstrating the intactness of the SAI in PD cases in the ''off'' condition.
Cortical cholinergic deficiency is more extensive and spreads to anterior cortical areas in PD patients advanced to PDD. 23 Therefore, we expected SAI to be abnormal in these patients. In accord with our hypothesis, we found that SAI is markedly reduced in PDD subjects. Manganelli et al. demonstrated that SAI became abnormal in nondemented PD patients in the presence of VH. 24 They reasoned that their findings may have a relation with an impending cognitive impairment. We demonstrated here that abnormality of SAI in PD subjects occurs when they are demented, irrespective of the presence of VH. Moreover, we have demonstrated that the magnitude of SAI response is highly correlated with the level of cognitive functioning.
Dopaminergic or GABAergic medications can also modulate SAI in healthy subjects and in AD, PD, and restless legs syndrome patients. 22,[25][26][27] Interaction of dopamine and acetylcholine in multiple sites may be responsible for the dopaminergic modulation of SAI. 25 Dopamine can increase acetylcholine release from cortical cholinergic terminals or cholinergic interneurons. 28 Also, dopaminergic stimulation of nucleus accumbens can change the activity of nucleus basalis of Meynert neurons through its GABAergic projections to the basal forebrain. 29 We speculate that this modulation can be even more overt by nonphysiological dopaminergic stimulation in dopamine-and/or acetylcholine-deficient states, such as PD, PDD, and AD. Furthermore, alteration in multiple neurotransmitters occurs in these disorders. Thus, the possible contribution of altered states of other neurotransmitters cannot be excluded.
These findings show that differential cholinergic deficiency occurs in PD and PDD: Functional cholinergic deficiency in frontal cortical areas (e.g., sensorimotor cortex) occurs in PDD, but not in PD. It can be a neurophysiological correlate of PDD. Furthermore, because in AD, recovery of SAI after a single dose of acetylcholinesterase inhibitor was reported to predict long-term treatment response, 30,31 it may also be used in the follow-up of treatment response in PDD patients, together with cognitive profiling. 9

A B S T R A C T
Background: Many factors can jeopardize the accuracy of deep brain stimulation (DBS) lead placement. Confirmation of lead placement while the patient is still in the operating room would be advantageous. Intraoperative MRI or CT can identify placement errors, but these modalities can be cost-or time prohibitive. Intraoperative fluoroscopy may give information on the accuracy of the Y coordinate, but the accuracy of the X coordinate usually cannot be confirmed. When an object of known dimensions is present in the brain, such as a unilateral DBS lead, its dimensions can be used to calculate unknown distances. The objective of this study was to determine if intraoperative AP skull x-ray accurately predicts the distance between DBS electrodes using postoperative MRI as the gold standard.
Methods: The distance between 32 pairs of DBS leads was measured by 2 independent raters under blinded conditions on intraoperative AP x-ray and postoperative axial and coronal MRI. Variable x-ray magnification was accounted for using the formula: actual distance between 2 leads 5 (measured distance between DBS leads)/ (average measured length of electrodes) 3 7.5 mm.
Results: The mean (6 SD The stereotactic nature of deep brain stimulation (DBS) procedures precludes direct visualization of microelectrodes and DBS leads intraoperatively. Although these probes usually follow their intended course toward the target, this may not always be the case. Errors due to frame problems, MRI distortion, incorrect targeting, brain shift, or electrode deviation can diminish the accuracy of stereotactic surgery. Intraoperative imaging with MRI or CT scans would identify these errors while still correctable but may be impractical because of cost, time, and space restraints in the operating room (OR). 1,2 Portable AP skull x-rays can visualize micro-as well as DBS electrodes, but using x-ray to measure intracranial objects is challenging. First, there is an unknown magnification factor that varies with both the distance from the camera to the object (the farther away the camera, the less the magnification) and the film to the object (the closer the film, the less the magnification). Second, the x-ray beam needs to be rectilinear to the object of interest and the film in order to prevent underestimation of the size of the object.
In the era of stereotactic lesioning, when stereotactic x-ray was the only imaging method available to neurosurgeons, this was dealt with by placing the x-ray cameras far away (3.5 m) from the frame center to minimize magnification and by placing the cameras in a fixed, orthogonal position relative to the center of the frame, usually in the wall and ceiling of the OR. 3,4 Nowadays CT-and MRI-based targeting has largely obviated the need for stereotactic x-ray, and only few ORs still have the built-in x-ray equipment; in addition, the position of the OR table has become variable from one surgery to the next. Therefore, intraoperative x-ray has largely lost its utility. 5 With DBS largely replacing lesioning, however, the situation has become quite different because now a reference object of known size (the DBS lead) is incorporated in the area of interest. This scenario presents itself during placement of the second lead of a bilateral procedure, where the first lead with its known dimensions provides such a reference object. Movement of microelectrodes in the mediolateral plane and position of the second DBS electrode could then be determined relative to the first DBS lead. This method would be most valuable in cases in which the distance of the initial, or ''reference,'' DBS lead to the midline is known from preoperative MRI, as would be the case in staged bilateral procedures or when performing DBS lead revisions. However, the accuracy of this method is completely unknown.
In the current feasibility study, we tested the hypothesis that intraoperative x-ray can accurately and reliably predict the distance between DBS electrodes using postoperative MRI as the gold standard. If confirmed, we believe that simple x-ray will provide useful intraoperative imaging information about microelectrode or lead positioning.

Patients and Methods
We included imaging data of 32 consecutive Parkinson's disease (PD) patients who underwent either contemporaneous or staged bilateral STN DBS and who had a complete and technically adequate imaging set (intraoperative AP x-ray, postoperative axial MRI, and postoperative coronal MRI) available for review. Intraoperative x-rays were taken under standardized conditions (Fig. 1). The film cassette was held by the surgeon in a sterile plastic cover immediately behind the patient's head, at an angle parallel to the arc of the frame and therefore parallel to the drive with the microelectrode or DBS lead. The face of the camera was also positioned parallel to the arc at a standard distance for skull x-rays of about 40 inches. The x-ray parameters were standard, at 80 kilovolt peak (kVp) and 40-80 milliampere-seconds (mAs). For clinical measurement in the OR, the film was maximally magnified and printed. The image was also loaded into the PACS system from where study measurements were made offline.
Two independent raters measured the distances between DBS electrodes on all 3 imaging modalities using measurement tools in the PACS system. Each rater measured 1 modality at a time, without access to the other 2 modalities, their previous measurements, or the other rater's results.

AP X-Ray Measurements
On AP x-ray, the distance between the leads was measured at the level of the bottom of contact 1 (Fig. 2). The length of each DBS electrode (Medtronic 3389) was measured from the top of contact 3 to the base of contact 0, and the average was calculated. The actual size of the lead is 7.5 mm. Accounting for differences in x-ray magnification, the actual distance between the 2 leads was then calculated with the following formula: actual distance between 2 leads ¼ (measured distance between DBS leads)/(average measured length of electrodes) Â 7.5 mm (if the widerspaced electrode [Medtronic 3387] is used, replace 7.5 by 10.5 in the formula).

Postoperative MRI Measurements
On axial MRI (T2-weighted, 2-mm contiguous slices). measurements were taken on the axial slice 2 mm above the bottom of the lead as defined as the lowest slice where any hypointensity could be visualized as a best attempt to correlate with the location of contact 1.
On coronal MRI (T2-weighted, 2-mm contiguous slices). the distance between the bottom of the hypointensities corresponding to contacts 1 was measured. If the 2 leads were not in the same coronal slice, the distance of each lead to the midline was measured and added up.
Data were analyzed using 2-way analysis of variance (ANOVA) to detect differences between the 3 imaging modalities and between the 2 raters and to examine a modality Â rater interaction. The single measure intraclass correlation coefficient (ICC) was used to examine interrater variability/reliability, and a 1-sample t test was used to test if the differences in measurement between the 2 methods were significant.

Results
Of the 32 cases analyzed, 24 were contemporaneous bilateral STN procedures, and 8 were staged bilateral STN procedures. Intraoperative skull x-ray and postoperative axial MRI were available in all 32 cases, whereas coronal MRI was available in 30 of 32 cases.
Distances between DBS leads are shown for each modality and each rater in Table 1. The mean (6 SD) distance on x-ray was 22.62 6 2.23 mm, on axial MRI 22.78 6 1.90 mm, and on coronal MRI 22.79 6 2.00 mm. ANOVA revealed no difference based on method (P ¼ .887) or rater (P ¼ .940). In addition, no interaction was found between methods and raters (P ¼ .964).
Using the averaged measurements of the 2 raters, the difference between x-ray and axial MRI method was (mean 6 SD of the absolute differences) 0.85 6 0.68 mm; between x-ray and coronal MRI methods, 0.86 6 0.77 mm (range, 0.01-2.60 mm); and between axial MRI and coronal MRI method, 0.41 6 0.44 mm (range, 0.01-1.74 mm).
There was more frequent underestimation of distance using x-ray compared with axial MRI (21 of 32 measurements) and compared with coronal MRI (21 of 30 measurements). The differences in measurement between the 2 methods were not significantly different from zero for  3 (a, b, c) are in the same plane and parallel to one another. d: Distance between the film cassette and the electrode was kept as small as possible. During imaging the film cassette was held in optimal position by a member of the neurosurgical team. e: Distance between frame and camera was about 100 cm. f: For optimal contrast and density, 80 kilovolt peak (kVp) and 40-80 milliamp-seconds (mAs) were used. x-ray/axial MRI (1-sample t 31 ¼ 0.80, P ¼ .43) and x-ray/coronal MRI (1-sample t 31 ¼ 1.23, P ¼ . 23) Linear regression showed strong correlations between measurements on x-ray and axial MRI, with an R 2 coefficient of 0.76. Similarly, strong correlations existed between measurements on x-ray and coronal MRI, with an R 2 coefficient of 0.74, as well as between measurements on axial and coronal MRI, with an R 2 coefficient of 0.91.

Discussion
Our findings demonstrate that AP x-ray can provide reliable intraoperative information regarding electrode location with submillimetric accuracy in the mediolateral plane.
The obvious limitation of the method is that it can only be used with 1 lead already in place, as it depends on having an object with known dimensions (ie, the DBS lead of 7.5 or 10.5 mm) in the x-ray image to correct for the variable magnification inherent to portable x-ray. In addition, it is important to ascertain that the x-ray beam is orthogonal to the lead and the film in order to prevent parallax. Given the proximity of the leads to the midline, however, the effects of mild rotation of the head in the frame are small. Several scenarios exist where this method can be of value to the surgical team.

Scenario 1: Staged Bilateral DBS
AP x-ray is most useful when performing a staged bilateral procedure because the distance between the original DBS lead to the midline will be known from the stereotactic MRI obtained prior to the second, contralateral surgery. Subtraction of this distance from the interelectrode distance on intraoperative x-ray will result in an accurate estimation of the distance between the second electrode (microelectrode or DBS lead) and the midline.
If MER is used and multiple microelectrode tracts are made, AP x-ray can be useful to confirm that the intended microelectrode movement in the mediolateral plane has indeed been realized. Using this technique, we have documented that electrodes sometimes move too little, too much, or not at all. If undetected, this may lead to misinterpretation of the neurophysiologic findings and, as a result, to unnecessary MER tracts or suboptimal DBS lead placement. Similarly, if during test stimulation through the DBS lead, the therapeutic window turns out to be too narrow and repositioning is necessary, the lead may inadvertently follow the path of least resistance and slip into the previous tract. Although in the alert patient this may be deducted from identical stimulation thresholds for benefits and side effects as before, this may not always be the case, and confirmation that the lead did not move with AP x-ray will support the decision to reposition the lead once more.

Scenario 2: Contemporaneous Bilateral DBS
Even when the distance between the first lead and the midline is not known, as is the case with contemporaneous bilateral procedures, we still find it useful to obtain AP x-rays once we start working on the second side. Microelectrode mapping and test stimulation through the first DBS lead will tell us whether that lead is a bit lateral, medial, or ''just right.'' If subsequent contralateral microelectrode recording or test stimulation leads to unexpected findings, we measure the interelectrode distance to rule out major deviations medially or laterally from the intended target.

Scenario 3: Lead Replacement
Another scenario where AP x-ray is extremely useful is when replacing a malpositioned DBS lead. The new lead is placed with the old lead still in situ, providing an opportunity to confirm the accuracy of the placement of the new lead in the mediolateral plane.
Our findings suggest that this low-cost and widely available method provides reliable intraoperative information regarding electrode location in the mediolateral plane. We have incorporated intraoperative AP x-ray as a standard procedure during the second arm of bilateral staged or contemporaneous DBS procedures to confirm the relative location of the microelectrode or DBS lead as well as during lead-repositioning procedures.

A B S T R A C T
Background: Rett syndrome (RTT) and autism disorder (AD) are 2 neurodevelopmental disorders of early life that share phenotypic features, one being hand stereotypies. Distinguishing RTT from AD often represents a challenge, and given their distinct long-term prognoses, this issue may have far-reaching implications. With the advances in genetic testing, the contribution of clinical manifestations in distinguishing RTT from AD has been overlooked.
Methods: A comparison of hand stereotypies in 20 children with RTT and 20 with AD was performed using detailed analyses of videotaped standardized observations. Results: Striking differences are observed between RTT and AD children. In RTT, hand stereotypies are predominantly complex, continuous, localized to the body midline, and involving mouthing. Conversely, in AD children, hand stereotypies are simple, bilateral, intermittent, and often involving objects. Conclusions: These results provide important clinical signs useful to the differential diagnosis of RTT versus AD, especially when genetic testing for RTT is not an option. Rett syndrome (RTT) and autism disorder (AD) are 2 sporadic neurodevelopmental disorders of early life; even though 75%-95% of RTT cases are linked to MECP2 mutations 1,2 and AD has a strong polygenetic basis. Although they represent 2 distinct neurological conditions, RTT and AD, especially in severely affected children, share clinical features such as poor sociability and lack of communication, along with irritability and anxiety. 3 Furthermore, in addition to their signature ''handwashing'' stereotypies, girls with RTT have many other stereotypies, like flapping and pacing, which are also observed in children with AD. 4 Conversely, hand stereotypies are, in fact, far from specific to RTT, as they can be observed rather often in children with AD. 5 These observations sometimes render the distinction between RTT and AD quite challenging for practitioners, and because RTT is ultimately fatal whereas AD is not, such a differential diagnosis is critically important.
Here, we report striking differences in the characteristics of hand stereotypies in children with RTT compared with those of matched cognitively impaired children with AD. We posit that these differences afford invaluable clues to distinguishing between RTT and AD on a clinical ground, especially in those situations where MECP2 genetic testing is not readily accessible.

Subjects
The 20 girls with RTT fulfilled revised criteria for RTT 6 and were examined and videotaped by the same child neurologist in Portugal. All children were carriers of an MECP2 mutation and had a NVIQ < 50; 9 walked independently. Ten were medicated with an antiepileptic (sodium valproate or carbamazepine). Their mean and median chronological age was 60 months (range, 36-96 months).
The 20 low-functioning children (11 boys, 9 girls) from the United States had a preschool diagnosis of autistic disorder based on DSM-III-R criteria. The number of available autistic girls who met criteria for this study was insufficient to limit selection to girls, but our previous study 5 and that of another group 7 showed no sex differences in stereotypies. The children were participants in a large multicenter longitudinal study of developmental disorders recruited between 1985 and 1988. 8 They were selected randomly from those in our previous study 5 who presented stereotypies and were matched pairwise on chronological age (mean, 68 months; range, 33-98 months) and NVIQ (<70; mean, 31; median, 27) with the girls with RTT. All were ambulatory, and none had a frank sensorimotor deficit or known neurological disorder; none was receiving high doses of anticonvulsant or psychotropic medication.
The parents of all the children signed an informed consent for participation and videotaping. The parents of the children presented here signed an additional consent for publication of their child's video and photograph.

Scoring
We independently scored a 5-minute consecutive video segment selected to be most representative of the child's overall behavior. We focused on hand stereotypies and developed a coding system to record the type of hand movement (eg, clapping, clenching), the position of the hand (joined or apart), their localization (eg, midline or away from the body), their laterality (eg, bilateral or unilateral), and their complexity 9 (simple stereotypies were defined as single movement involving 1 group of muscles [eg, handclapping], whereas complex stereotypies were defined as clusters of different coordinated movements performed in the same sequence and involving a group of muscles [eg, opening and closing the hands with finger extension]). 10 To be counted, a stereotypy had to be seen at least twice in order to be able to document its repetitiveness. The characterization and scoring of each stereotypy are presented in Table 1. The inconsistent filming of the face because of the fixed camera or the focus on the whole body did not allow for reliable scoring of associated facial movements. However, the videotapes provided enough details to allow for the coding of hand movements directed toward the mouth (ie, mouthing). The 2 trained authors coded each child's videos in both groups independently, with an interrater reliability kappa !0.8. A third coder, blinded to diagnoses, scored 70% of the entire sample, with an interrater reliability kappa ranging from 0.8 to 1.0 for the 9 variables. Scores in the 2 groups were compared using chi-square analysis.

Results
In the present study, 40 children with hand stereotypies are surveyed using videotape analyses. Among these, 20 had a confirmed diagnosis of RTT, and 20 had a definite diagnosis of AD. In this cohort of 40, hand stereotypies were associated with rocking, pacing, and skipping in 25% of AD children and in none of RTT children. All 40 children exhibited more than 1 type of hand stereotypy, yet RTT children typically displayed a much greater variety of hand stereotypies than did AD children (see Table  1). With respect to the complexity of the stereotypies, in the AD group, they appeared as simple hyperkinetic movements such as finger wiggling (40%) and hand flapping (30%). Complex hand stereotypies were observed in both groups, but with some salient differences. For instance, hand washing was only observed in RTT children, whereas shaking or tapping objects was only observed in AD. Furthermore, even when RTT and AD children exhibited the same type of complex hand stereotypies, they often differed in quality and frequency. In particular, clapping in RTT appears slow and monotonic, with forceful holding of hands together, whereas in AD children, clapping appears rapid and variable in its presentation, and when the hands are together, they touch briefly and lightly (see Video 1). Our videotape scoring revealed other striking and distinct features. During the selected 5-minute observation that started with the occurrence of hand stereotypies, all the RTT children had continuous hand movements, whereas all the AD children had intermittent clusters of repetitive hand movements (see Table 2). Furthermore, hand stereotypies in RTT children were localized predominantly in front of and at a close distance to the midline of the chest. In 60%, they were accompanied by mouthing, and in 70%, hands were joined or close to each other (see Video 2). Conversely, in the AD group, the repetitive patterned movements occurred in 75% away from the body with hands apart. Unilateral hand stereotypy was 2.5-fold more frequent in RTT than in AD children, whereas bilateral hand stereotypy occurred equally in both groups. Hand gaze was never observed in this RTT group, whereas 20% of children with AD exhibited close inspection of fingers or objects (see Video 3).

Discussion
Our video scoring revealed prominent differences in the phenomenology of stereotypies between RTT and AD. In RTT, hand stereotypies were continuous and predominated at the midline, whereas in AD, these stereotypies were intermittent and away from the body. In addition, when stereotypy episodes involved objects or hand gaze, the child always belonged to the AD group, never to the RTT group. Our observation is consistent with the findings from a recent study of 144 individuals with RTT. 11 In this study using family videos-a less systematic method-the authors also reported that repetitive movements in RTT were often continuous, localized in the midline of the body, and bilateral.
The phenotypic overlap between RTT and AD has long been recognized. 12 However, findings such as neurodevelopmental regression after a normal period of development, a loss of hand skills along with the appearance of stereotypies, deceleration of head growth, and progressive motor deterioration help to support the diagnosis of RTT. Our study has demonstrated that a fine clinical examination of hand stereotypies can reveal straightforward features that may further strengthen the diagnosis of RTT, hence helping in the differentiation of RTT from AD. Although the differential diagnosis of RTT versus AD can be achieved by genetic testing for MECP2 mutations, when such testing is either not available or not affordable, the clinical characteristics discussed above may be indispensable in distinguishing RTT from AD.
Thus far, little is known about the pathophysiology of hand stereotypies. However, that these repetitive movements occur in both RTT and AD suggests they may share a common neurophysiological basis, even if, as discussed above, their frequency and topology differ. Hyperkinetic movement disorders consistently point toward dysregulation in the basal ganglia circuitry, but the actual neuronal pathways implicated in motor stereotypies in general, and in hand stereotypies in particular, remain to be established. However, at the cellular level, we know that the timing of the regression period in RTT parallels the period of intense synaptic development. 13 Furthermore, neuropathological findings in RTT show selective reduction of dendritic spines in the pyramidal cells of the brain; this has also been reported in autism. 14 Thus, it may be suggested that the shared occurrence of stereotypies in RTT and AD results from comparable impaired integration of the basal ganglia/ motor thalamus input to the upper motor neurons because of failure of synaptic maintenance.
It is also worth mentioning that stereotypies in children have been suggested to be part of normal development 15,16 and that a developmental ''arrest'' might leave these children in a state where they are (physiologically) prone to stereotypic behavior. Although provocative at this point, this concept is important to consider but remains to be experimentally proven. Nonetheless, in our particular study, the repetitive movements we are reporting here are phenotypically dissimilar to those previously reported in normally developing children. 17,18 Legends to the Videos Video 1: Segment 1: A 5-year-old girl with Rett syndrome exhibiting continuous forceful clapping close to her body. Segment 2: A 3-year-old girl with autism disorder showing light rapid clapping.
Video 2: A four-year-old girl with Rett syndrome sitting on her mother's lap presenting with complex hand stereotypies associated with mouthing.
Video 3: A six-year-old boy with autism disorder scrutinizing toys. Pallidal deep brain stimulation (DBS) is recognized as an effective therapy in dystonia. 1 More recently, the subthalamic nucleus (STN), which is widely used for DBS in Parkinson's disease (PD), 2 has been described as an alternative DBS target for patients with dystonia. 3 Thus, hyperkinesia in dystonia and bradykinesia in PD are effectively treated by DBS in the same target areas of the basal ganglia (BG). It has been suggested that the diverse therapeutic effects of DBS may arise from the suppression of phenotype-specific pathologically enhanced oscillatory activity in the cortex-BG network, 4 such as the enhanced oscillatory beta activity seen in PD. 5 In patients with dystonia, enhanced oscillatory activity in the low-frequency band (4-12 Hz) has been recorded from the globus pallidus internus (GPi). 6,7 Such synchronization correlates and is coherent with EMG activity during involuntary (mainly phasic) dystonic muscle contractions, [8][9][10] suggesting it may contribute to the pathophysiology of dystonia. However, deep brain recordings of local field potential (LFP) activity in dystonia have so far been restricted to the GPi. Further support for the hypothesis that low-frequency activity plays a pivotal role in the generation of abnormal dystonic movements comes from LFP recordings of enhanced lowfrequency oscillatory activity in different nuclei of the cortex-BG motor network. Here, we present for the first time LFP recordings from the STN in a patient undergoing STN-DBS for severe generalized dystonia; the oscillatory pattern in this patient was compared with results from STN-LFP activity obtained in PD patients and from GPi-LFP activity in dystonic patients. Our results provide further evidence for phenotype-specific pathologically increased low-frequency activity in the cortex-BG network related to dystonic muscle activity.  LFP rest recordings from 5 PD patients off medication undergoing STN-DBS and 5 dystonia patients undergoing GPi-DBS from 2 archival data sets were included in the study. Clinical details have been reported previously. 11,12 Correct placement of the DBS electrodes was confirmed by intraoperative microelectrode recordings, direct macrostimulation, and postoperative MRI in all patients.

Recordings
All patients were studied at rest 2-5 days after implantation of DBS electrodes before their connection to the pulse generator. Written informed consent was obtained, and LFP recordings were approved by the local ethics committee. In our patient with dystonia, STN-LFPs were recorded while GPi-DBS was switched off. Recordings included periods of spontaneous dystonic posturing and involuntary movements. STN-LFPs were obtained from the 3 adjacent bipolar contact pairs (01, 12, 23) of each macroelectrode, amplified (Â50,000), and filtered at 1-250 Hz using a D360 amplifier (Digitimer, Hertfordshire, UK), and recorded at a sampling rate of 1 kHz through a 1401 A-D converter (CED, Cambridge, UK) onto a computer using Spike2 software. Simultaneous surface EMG recordings were made over the sternocleidomastoid muscle (SCM) bilaterally using pairs of Ag/AgCl electrodes, amplified (Â1000) and recorded using the same settings as for LFP recordings.

Analysis
All data were imported and analysed in Matlab (Mathworks, Natick, MA). FFT -based power spectra were calculated from 2 to 4 minutes of rest recordings in all patients. Power spectra and coherence estimates were computed with a segment length of 1024 samples, resulting in a frequency resolution of 0.98 Hz. Spectral power was expressed as the percentage of total power in the 4-to 95-Hz range (excluding 45-55 Hz) for each contact pair. In addition, mean STN-LFP power (recording duration of 182 seconds) in the single dystonia patient was calculated across contact pairs and sides and compared with the mean power from selected best-contact pairs of GPi-LFPs from 5 patients with dystonia (recording duration, 161 6 71 seconds [mean 6 SD]) and STN-LFPs in 5 PD patients off medication (mean recording duration, 268 6 67 seconds). Power at frequencies less than 4 Hz was excluded to minimize the contribution of movement artifacts.
In the dystonia patient, coherence estimates were calculated between the STN contact pairs with the highest power at low frequency and the SCM-EMG separately for ipsi-and contralateral muscles, with significance thresholds calculated according to Halliday et al 13 as upper 95% confidence limits (a ¼ 0.05). To evaluate directional influences between STN-LFPs and EMG, we calculated the partial directed coherence (PDC). PDC is based on Granger causality 14 and is computed from the coefficients of the multivariate autoregressive model transformed to the frequency domain. 15 The 95% confidence interval was computed based on the leave-one-out method. 16

Results
Distinct peaks in the power spectra were only observed in the low-frequency band at about 7 Hz in STN-LFPs from all contact pairs in our dystonia patient (Fig. 1A,B). For the right STN, low-frequency activity (4-12 Hz) showed a local maximum at contact pair 01 in line with the localization of those contacts within the STN. This oscillatory pattern was similar to the enhanced low-frequency activity that occurred in pallidal LFPs in the 5 patients with dystonia. In contrast, STN-LFPs in our PD patients off medication showed predominant beta (13-30 Hz) activity, with a peak at about 18 Hz (Fig. 1A-D), which is in line with previous findings. 5,7 Significant coherence between STN-LFPs and EMG was also found in the patient with dystonia in the low-frequency range (4)(5)(6)(7)(8)(9)(10)(11)(12). This was larger for STN-LFPs to the contralateral SCM-EMG compared with between STN-LFPs and the ipsilateral EMG ( Fig.  2A,B). Interestingly, the highest coherence peak was observed between right STN-LFPs and the left SCM-EMG, the latter associated with the muscle showing predominant dystonic activity during involuntary head movements. Directionality analysis revealed bidirectional information flow at low frequency, with predominant drive from the right STN to the left EMG (Fig. 2C).

Discussion
We have shown here for the first time that enhanced low-frequency activity (4-12 Hz) is present in the STN in dystonia. The oscillatory STN-LFP pattern in our dystonia patient is clearly different from that in PD patients, which extends previous results obtained from pallidal recordings in dystonia and PD. 6 In line with this, the STN and GPi both showed abnormal bursting activity during intraoperative unit recordings associated with dystonia. 17,18 In our patient, low-frequency STN activity was coherent with EMG dystonic spasms, with directionality analysis suggesting that STN activity might drive muscle in this frequency range. This is similar to what has been shown for low-frequency pallidal activity. 9 Our results support the hypothesis that disease-specific low-frequency activity is enhanced in the cortex-BG network in dystonia and may contribute to the occurrence of abnormal movements. Although our data provide only correlative evidence and do not indicate where in the cortex-BG loop this pathological activity is generated, they do suggest that either nucleus (STN or GPi) can be targeted to interrupt pathological activity and prevent its further propagation within the cortex-BG network. 4,5 This might provide an explanation for why STN-DBS is also effective in patients with dystonia. Moreover, it might explain why there was no significant additional clinical benefit of combined STN-and GPi-DBS in our patient. The limited clinical efficacy of DBS of both targets in our patient suggests that additional, possibly compensatory mechanisms or longterm plastic changes within the cortex-BG loop need to be established for clinical improvement. However, our observation is limited by our reporting results from a single case only. Further studies are thus required to evaluate the STN as an additional target for DBS in dystonia. Furthermore, it cannot be excluded that GPi-DBS might have induced long-lasting changes in BG activity. However, our results show remarkable similarity to previous findings in dystonia patients, thus supporting the hypothesis that low-frequency oscillations in the cortex-BG loop are an elec-trophysiological signature of dystonia and play a pivotal role in the generation of involuntary muscle contractions.