Clinical and demographic profile is provided in Table 1. Eight patients were males and three were females, with age at presentation ranging from 1 week to 12 years. Eight of them had unilateral clefts while three had bilateral involvement. Eight of these 14 hemispheres had closed-lip schizencephalies, while 6 had open-lip clefts. All 11 of them had intractable epilepsy. The median age at onset of seizures was 4 months (range: 0.3-144 months). Ten of them had high seizure frequency, ranging from 1-2/day to 20/day. Only 1 patient (#6) had infrequent seizure episodes, up to 1 in 3-4 months. Multiple seizure types were seen in conjunction in these patients, including complex partial seizures in 8 patients, infantile spasms in 6, some episodes of generalized tonic-clonic seizures in 3, and absence and myoclonic seizures in at least 2 patients. The median follow-up duration was 4 years (range: 0-17 years).

All of the patients were developmentally delayed, and most of them had associated findings, most common being contralateral hemiparesis, seen in 4 patients. Other common associated conditions included congenital hydrocephalus of unknown etiology in 2 patients (#1 and 6), skull abnormalities (including dolichocephaly, plagiocephaly and microcephaly) in 3 (#6, 8 and 11), scoliosis in 1 (#3), spastic cerebral palsy in 4 (#3, 6, 8 and 11), septo-optic dysplasia in 2 (#2 and 11), and other ocular abnormalities (esotropia) in 4 patients (#1, 6, 10 and 11).

We found a variety of possible etiologies in our patient cohort. One patient (#8) had a genetic abnormality: 16p11.2 duplication, which was suggested as a cause for the schizencephaly. This patient's mother had threatened preterm labor at 30 weeks, at which time the prenatal ultrasound also showed a single umbilical artery. Two patients had histories of infarcts in-utero; on MR angiography one (#10) showed decreased caliber of the left middle cerebral artery, with a hypoplastic left posterior communicating artery, and the other one (#1) showed developmental absence of a segment of the right anterior cerebral artery. These babies had the schizencephalic clefts in the corresponding areas of distribution of these respective arteries. Another patient (#8), without overt history of infarct in-utero, showed asymmetry and splaying of vessels in the middle cerebral distribution on MRA, in the area of and on the same side as the malformation. One patient (#6) was found to have hydrocephalus at 20 weeks of gestation.

In all the patients, FDG-PET scans showed a larger area of abnormality than MRI (Figure 1), and corresponding better with the electrophysiological changes compared to MRI (Table 2). Also, the area of involvement on the FDG-PET scan showed a positive correlation with the extent of developmental delay; though a formal statistical analysis was not performed due to very small sample size.Five patients, with unilateral disease on MRI, underwent epilepsy surgery (4 hemispherectomy and 1 multilobar resection). Out of these, 2 patients (#5 and 8) had only focal defect on MRI, but both the video-EEG and FDG-PET showed a much larger area of involvement. They underwent hemispherectomy after confirmation with subdural electrode placement, and the widespread involvement was subsequently also corroborated by histopathological examination of the resected brain tissue which showed, in one patient, abnormal neocortical layering/ malformation of cortical development of a significant area of 2 lobes adjacent to the cleft and, in the other patient, scattered foci of vacuolation, myelin loss and gliosis in multiple lobes. One patient (#6) with a unilateral video-EEG abnormality and an equivocal MRI (with a unilateral porencephaly, and possible schizencephaly on the opposite side) was shown to have bilateral disease on PET and, therefore, was not operated. On the other hand, the only patient (#8) who was not seizure-free after surgery had unilateral schizencephaly and video-EEG changes, but bilateral abnormalities on FDG PET.

At last follow-up, six patients were seizure-free, but only 2 had discontinued their anti-epileptic medication. These patients continued to have seizures for an average duration of 26 months (range: 3.5-116 months) before achieving seizure- freedom. Seizures in 2 of these 10 patients responded to anti-epileptic medications, including valproic acid in one, and the combination of adrenocorticotropic hormone (ACTH), zonisamide and lamotrigene in the other (median seizure-free duration 56 months, range: 1-120 months). All 6 patients with infantile spasms were treated with ACTH, and four of them responded to it; however, the response was very brief, ranging from 1 month to 1 year, after which the seizures recurred in three of them and one was lost to follow-up. None of the patients was treated with vigabatrin. Five patients underwent epilepsy surgery, after being tried on at least 4 anti-epileptic drugs. Four of these 5 patients were seizure-free at last follow-up (average seizure-free duration: 37 months, range: 0.3-188 months).

Schizencephaly usually presents with developmental delay, seizures and motor deficits [10]. The prevalence of developmental delay in schizencephaly patients was found to be 83%. All of the children in our study had global developmental delay. The severity of the developmental delay depends on the extent of the malformation [10]. In previous studies, the mean age of seizure onset was found to be 13 months and the most common seizure type was complex partial seizures [12]. It has been reported that the prevalence of epilepsy in schizencephaly is 37-65% [1, 13]. However, the median age at seizure onset in our study population was much lower and all the children had intractable epilepsy, probably attributable to the tertiary care epilepsy surgery referral status of our center. Seizures have been reported to be worse in unilateral disease, as bilateral clefts have been hypothesized to inhibit the propagation of epileptiform discharges [10]; however, we found equally worse seizures in cases of bilateral involvement. Other clinical features (e.g., cognitive function) may be worse in bilateral disease due not only to the malformation, but also the lower capacity for cortical reorganization when both hemispheres are abnormal. General intellectual function and neurodevelopmental outcome was found to be dependent on the extent of cortical involvement as well as the presence of other CNS anomalies [14, 15]. Since PET scanning provides a functional assessment of the brain, it can be useful in delineating the extent and severity of the cortical involvement and eventual functional status of the child. We observed that in our cohort of patients, children with larger area of involvement on PET scan, inclusive of the structural abnormality, were more delayed as compared to those with less extensive involvement.

The most commonly accepted theory for the development of schizencephaly is vascular disruption in the developing fetal brain around the time of neuronal migration. In fact, the proposed underlying mechanism for viral and drug-induced etiologies of schizencephaly is also vascular disruption [6]. In the spectrum of polymicrogyria-porencephaly-schizencephaly- hydranencephaly, the exact anatomic lesion that will develop depends on the severity and timing of the insult that occurred during brain development [6]. An antenatal and postnatal color-flow Doppler imaging study found middle cerebral artery flow abnormality at 22-32 weeks of gestation, suggesting that early vascular occlusion can cause schizencephaly [16]. In our study, we found that two children had hypoplasia of an intracranial artery on MRA; one involved the posterior communicating artery, and the other one involved the anterior cerebral artery. The MRI scans of these children showed that their clefts were in the area of distribution of these respective arteries, suggesting that the vascular hypoplasia may be etiologically related to the cleft formation.

An interesting observation from our study is the poor response of children with schizencephaly to ACTH. Although infantile spasms often respond well to ACTH, in our patients with coexisting schizencephaly, the response to ACTH was either temporary or completely absent. This finding suggests that ACTH may not be the best choice in schizencephaly with infantile spasms. ACTH is an expensive drug, with multiple side effects including fluid retention, hypertension, bloating, and if its effect is limited in this population, other approaches may be attempted first, such as vigabatrin or ketogenic diet, which were not used in patients from this old series. This is a limitation of our study. However, larger studies are required to confirm that ACTH may not be the best choice.

Some neuronal migration defects are known to have underlying diffuse brain injury that develop during the same period [14]. Schizencephaly, whether unilateral or bilateral, is very often found to be associated with other defects, including cortical dysplasia, cerebral atrophy, and posterior fossa abnormalities. These abnormalities may be found anywhere, such as adjacent to the cleft, ipsilateral but remote, or contralateral in a symmetric or asymmetric distribution [1, 15, 17, 18]. Seizures usually originate from the grey matter surrounding the cleft, but they can also arise from any of the other abnormal areas in the cortex. We still do not have a complete understanding of the spatial correlation between the area of malformation and the epileptogenic zone [19]. MRI is considered the gold-standard for the diagnosis of schizencephaly. One of the drawbacks of MRI, however, is that it is sometimes unable to detect small, subtle defects [20, 21]. Kim et al. [22](19) suggested that the difference in the ability of the MRI and PET scans becomes marked when the defect is small (MRI was able to detect 13%, whereas PET scan could detect 86% of defects. Another problem is that since the grey matter lining the cleft may not always be the epileptogenic region, the EEG and MRI findings can be discordant in many schizencephaly cases [15]. Similarly, EEG, which is the diagnostic test of choice for identifying the epileptic focus, may provide false lateralization or localization in cases of large defects due to dipole distortion. This renders the EEG less useful in such cases and other imaging modalities have to be relied upon [20]. FDG PET can be particularly useful in these cases, as shown in our study.

In the FDG-PET scan, abnormal areas show up as either characteristic hypometabolism or as areas of hypermetabolism. Hypometabolic areas may represent underlying neuronal dysfunction related to several factors, such as absence of cortical matter in the cleft, microdysgenesis/dysplasia of the surrounding cortex or possible epileptogenic cortex, whereas hypermetabolism can be a more specific finding with several possible explanations. First, clinical or subclinical seizures during the tracer uptake period due to the intrinsic epileptogenicity of the dysplastic cortex can result in a hypermetabolic focus on the scan [23, 24, 25], and therefore the hypermetabolic area may represent the epileptogenic cortex, thus helping in seizure lateralization/localization. Continuous EEG monitoring during the FDG uptake period, as routinely done at our center, can be particularly helpful in these cases. Indeed, three of our patients showed focal/ generalized spike and wave activity with hypermetabolism on the PET scan. Second, associated heterotopic malformations, which are often present in the cortex of these patients can themselves be seen as hypermetabolic as a result of the dense accumulation of neurons and neuronal connections [11, 20, 26]. In addition, these heterotopias, can also cause functional reorganization of adjacent areas or even contralaterally [20, 26, 27]. Epilepsy surgery can be offered if the seizures become intractable and the epileptogenic zone can be identified. Although the occurrence and type of seizures are not correlated with the type of schizencephaly, it has been found that bilateral open lip schizencephalies commonly have an earlier age of onset and are refractory to multiple seizure medications [17]. However, children with bilateral clefts are not eligible for surgical treatment. There are two important pre-surgical considerations for a good surgical outcome, the first being the exact identification and demarcation of the epileptogenic zone that will be resected and the second, the evaluation of the remaining brain, to ensure that the post-surgical functional deficit will be acceptable. Here again, FDG-PET can be helpful, as it can identify the epileptogenic cortex, as well as provide information about the functional status of the remaining brain.

In our study, we found that FDG-PET scans always showed a larger area of involvement as compared to MRI, and the PET findings corresponded better with the EEG. This observation is in agreement with other studies indicating that the cortex beyond the cleft may also be functionally deficient [22]. We found that the histopathological examination of the resected brain tissue corroborated with the PET scan findings, showing malformation of cortical development or foci of vacuolation, gliosis and myelin loss, where no abnormality was detected on the MRI scan. This shows that PET scans can identify non-structural abnormalities and guide the decision of the eligibility of a particular candidate for surgery. In one patient, bilateral abnormality on PET scan helped in re-evaluation of an unequivocal unilateral MRI, precluding hemispherectomy in a child who would otherwise have had a poor surgical outcome as the remaining hemisphere would also have been abnormal. One of our patients who did have a hemispherectomy, despite bilateral functional abnormalities on PET scan, continued to have seizures (Figure 1). We note that ambiguous findings on MRI scans should be re-evaluated in the light of positive PET scan findings to detect subtle defects [20, 28].

Although pre-resection subdural grid electrode placement is the final and most accurate way to identify the epileptogenic region, it is an invasive procedure, which should be avoided if the patient has already been determined to be suboptimal for surgery through non-invasive tests [17]. We suggest that FDG-PET scans may be an adjunctive tool in the pre-surgical evaluation of schizencephaly, providing additional information, especially when MRI findings are either normal or discordant with EEG and before resorting to subdural grid placement [29].

Our study does have some limitations. First, patients with very small schizencephalic clefts may be under-represented in our tertiary center series, since they have mild symptoms and may not have been scanned for a cerebral abnormality, or may not have referred to a tertiary center. Second, the small sample size precludes making conclusive inferences, though this may be an unavoidable problem due to the rarity of this condition and associated logistic constraints.