Lesion Detection Through MRI Postprocessing in Pathology-Proven Focal Cortical Dysplasia: Experience at a Single Institution in the Republic of Korea

Abstract

Background and Purpose

Focal cortical dysplasia (FCD) is one of the most common causes of drug-resistant epilepsy, and necessitates a multimodal evaluation to ensure optimal surgical treatment. This study aimed to determine the supportive value of the morphometric analysis program (MAP) in detecting FCD using data from a single institution in Korea.

Methods

To develop a standard reference for the MAP, normal-looking MRIs by two scanners that are frequently used in this center were chosen. Patients with drug-resistant epilepsy and FCD after surgery were candidates for the analysis. The three-dimensional T1-weighted MRI scans of the patients were analyzed as test cases using the MAP.

Results

The MRI scans of 87 patients were included in the analysis. The radiologist detected abnormal findings correlated with FCD (RAD positive [RAD(+)]) in 34 cases (39.1%), while the MAP could detect FCD in 25.3% of cases. A combination of the MAP (MAP[+] cases) with interpretations by the radiologist increased the detection to 42.5% (37 cases). The lesion detection rate was not different according to the type of reference scanners except in one case. MAP(+)/RAD(-) presented in three cases, all of which had FCD type IIa. The detection rate was slightly higher using the same kind of scanner as a reference, but not significantly (35.0% vs. 22.4% p=0.26).

Conclusions

The results of postprocessing in the MAP for detecting FCD did not depend on the type of reference scanner, and the MAP was the strongest in detecting FCD IIa. We suggested that the MAP could be widely utilized without developing institutional standards and could become an effective tool for detecting FCD lesions.

INTRODUCTION

Focal cortical dysplasia (FCD) is a congenital disease that results from disruptions in neural proliferation and migration. It is one of the most common causes of drug-resistant epilepsy, which often necessitates presurgical evaluation to consider surgical resection as a treatment. The typical microscopy findings of FCD are cortical disorganization, bizarre and dysmorphic neurons, and balloon cells.¹ Before the confirmative diagnosis of FCD using pathology, radiologic diagnosis using MRI to identify the typical features, such as blurring of junctions between gray matter (GM) and white matter (WM), abnormal cortical thickening, or ‘transmantle sign,’ is essential for the decision-making and planning of surgical treatment.² However, even with recent developments in MRI techniques, FCD sometimes appears negative on MRI, which impedes its diagnosis.³

The ability to detect FCD in a presurgical evaluation is crucial because postoperative outcomes (seizure-free rate, seizure frequency, and the intensity of seizure semiology) are better in MRI-positive patients,⁴,⁵ and inaccurate or incomplete localization of the lesion for surgical resection might lead to seizure recurrence after postoperative cessation of antiseizure medication.⁶ Many efforts have been made recently to improve those outcomes and visualization techniques. The morphometric analysis program (MAP; version 2018 is MAP18, https://www.translationalneuroimaging.de/map18) has been tested and validated in multiple studies performed in Europe, the USA, and China, and has become a supportive MRI postprocessing tool focusing on FCD lesions⁷ since it was first introduced in 2005.⁸,⁹,¹⁰ The MAP is based on algorithms from the specialized software program for statistical mapping and compares brain images of the test case with those of the normal reference database.¹¹ Highlighting and visualizing a lesion suspected of FCD helps clinicians to localize epileptogenic foci and to make further decisions for treatment options.¹²

This study aimed to establish an institution-specific standard database for the MAP and present its diagnostic value as a supporting tool for existing multimodal presurgical evaluations in data from a single institution in Korea.

METHODS

Collection of the standard reference data set

We retrospectively collected epilepsy-protocol MRI data, including three-dimensional (3D) T1-weighted images, that were obtained from patients with epilepsy or with suspected seizures. Between 2009 (when epilepsy-protocol MRI scans were introduced) and 2020, 1,375 3D T1-weighted MRI examinations were performed with 8 different scanners. To develop a standard reference for the MAP specific to our center, MRIs by two frequently used scanners (Philips Ingenia CX 3-tesla and Siemens Skyra 3-tesla; www.usa.philips.com/healthcare/solutions/magnetic-resonance, https://www.siemens-healthineers.com/magnetic-resonance-imaging) were chosen among 965 normal-looking MRI scans, and 3D T1-weighted images from each scanner served as the inputs for the MAP analysis (Fig. 1).¹¹ The Ingenia group consisted of 101 patients (mean age of 34 years, age range of 16–69 years; including 58 males), while the Skyra group consisted of 93 patients (median age of 35 years, age range of 16–65 years; including 45 males). The normal reference of our institution comprising two scanners was incorporated into the MAP, which was kindly provided by the developer (Professor Hans-Jürgen Huppertz).

Fig. 1. Flowchart of the establishment of a normal MRI database from two different scanners and processing of the test cases. 3-dimensional T1-weighted images of the epilepsy-protocol MRI scans of 1,375 patients with epilepsy or with suspected seizures were retrospectively reviewed, and 965 normal-looking MRI scans were selected to construct the normal reference database. The data set was finally obtained from two representative scanners (Philips Ingenia CX 3-tesla and Siemens Skyra 3-tesla) among eight scanners. After pre- and postprocessing of 87 test cases compared with the normal reference database of each representative scanner using SPM5 and the MAP, respectively, the junction, extension, and thickness images were developed. FCD, focal cortical dysplasia; MAP, morphometric analysis program.

Patient selection

The candidates for the retrospective analysis were patients who were diagnosed with drug-resistant neocortical focal epilepsy (in any of the temporal, frontal, parietal, or occipital lobes), underwent surgical resection of focal epileptogenic lesions based on decisions by experts in various disciplines (neurologists, radiologists, nuclear radiologists, and neurosurgeons), and were diagnosed with pathology-confirmed FCD between September 1, 2009 and May 1, 2020. The MRI scans of the patients were collected to be used as test cases for the MAP and clinical data were reviewed. The MRI scans were performed before the surgery using the epilepsy protocol, which consisted of a 3D T1-weighted sequence, T2-weighted sequences with axial and oblique coronal views, and a T2-weighted FLAIR sequence with an oblique coronal view. MRI scans with or without contrast enhancement were included in this study. As a routine part of the multidisciplinary discussion to decide epileptogenic focus, an experienced radiologist read the MRI scans with full protocol and found any significant findings related to the focus. In our center, MRI with the epilepsy protocol was designated to one of the ten scanners. The MRI parameters for each scanner are summarized in Supplementary Table 1 (in the online-only Data Supplement).

MRI postprocessing

The MAP18 was run via MATLAB (version R2019b, Math-Works) on 3D T1-weighted images of each test case. The preand postprocessing of 3D T1-weighted MRI scans were performed using SPM5 software as described previously.¹³,¹⁴ Preprocessing consisted of two steps: 1) normalization and correction of intensity and 2) simultaneous segmentation based on different brain compartments (GM, WM, and cerebrospinal fluid). Based on the radiologic features of FCD, the MAP18 computed the junction (to estimate the extent of blurring at the GM-WM junction), extension (to evaluate abnormal extensions of GM into WM), and thickness (to determine any abnormal thickness in the cortical ribbon).

The junction image was created next after the preprocessing. Individual upper and lower intensity thresholds for filtering and conversion to a binary image were determined based on the means and standard deviations of the voxel intensities in the GM and WM compartments. To create the extension image, the GM image was smoothed using a Gaussian kernel with an isotropic full width at half maximum of 6 mm. The smoothed GM image of the normal database was subtracted voxel by voxel from the GM image of the test case for comparison. For the thickness image, conversion to a binary image was performed with a cutoff value of 0.5. The run-length vectors were determined to estimate the cortical thickness of each voxel within the GM compartment, and the mean run-length images of the test case and the normal database were compared. MAP-positive criteria were determined through multiple steps as described previously.¹³,¹⁴ The reviewer searched for the bright focal area in each image, which indicated a significant deviation from the normal value (high z score), in the order of the junction, extension, and thickness images because the junction image is known to be the most-sensitive map. The z-score thresholds for the MAP-positive (MAP[+]) candidate were retrospectively determined to be 4 for the junction and thickness and 6 for the extension map. Naїve MRI scans without postprocessing (3D T1-weighted and FLAIR sequences) were then reviewed, and MAP(+) was finally confirmed if the abnormality was also found in the candidate lesion revealed by the MAP.

The results from the MAP18 analysis were divided into four categories: RAD(+)/MAP(+), RAD(+)/MAP(-), RAD(-)/MAP(+), and RAD(-)/MAP(-). The term “RAD” was used to classify the result of the traditional radiologist report as positive or negative, where RAD(-) meant that the MRI was initially interpreted as “normal” by the radiologist, while RAD(+) meant that the radiologist found a potentially epileptogenic lesion on the MRI scan. MAP(+) signified the cases in which the MAP detected FCD lesions, while MAP(-) signified no such detection; for example, RAD(+)/MAP(+) signified that the radiologist had identified an abnormal MRI finding, while the MAP had also detected FCD lesions. This study was approved by the Seoul National University Hospital Institutional Review Board (IRB No. 2106-106-1227) and followed the principles of the Declaration of Helsinki.

Statistical analyses

Statistical analyses were performed using SAS software (version 9.4) for Windows (SAS Institute, Cary, NC, USA). The χ² test was used for comparisons of categorical data, while Fisher’s exact test was used if values with expected frequencies of less than 5 comprised more than 20%. A probability value of p<0.05 was considered significant.

RESULTS

The MRI scans of 87 patients from 10 types of scanner (1 1-tesla, 3 1.5-tesla, and 6 3-tesla scanners) were finally enrolled into the retrospective analysis. Forty-nine patients were male. The median age of epilepsy onset was 13 (range 0–48 years). Most patients had temporal or frontal lobe epilepsy (73.6%). Medial temporal lobe epilepsy diagnoses were excluded. Table 1 lists other important clinical findings, including FCD subtypes and surgical outcomes categorized based on the International League Against Epilepsy (ILAE) classification.

Table 1. Clinical characteristics and demographic features of the 87 study patients.

Sex, male		49
Age at epilepsy onset, years		13 (0–48)
Age at MRI, years		28 (15–56)
Age at surgery, years		28 (16–58)
FCD subtype*
	I	57
	II	19
	III	14
Location
	Temporal	30
	Frontal	34
	Parietal	5
	Occipital	15
	Other or multifocal	3
Outcome: last follow-up after surgery (ILAE†)
	Class 1	35
	Class 2	8
	Class 3	9
	Class 4	24
	Class 5	9
	Class 6	2

Data are n or median (range) values.

^*Dual pathology of FCD subtypes were included; ^†Class 1, completely seizure free with no auras; class 2, only auras with no other seizures; class 3, one to three seizure days per year, with or without auras; class 4, four seizure days per year to 50% reduction from baseline seizure days, with or without auras; class 5, less than 50% reduction from baseline seizure days to 100% increase from baseline seizure days, with or without auras; class 6: greater than 100% increase from baseline seizure days, with or without auras.

FCD, focal cortical dysplasia; ILAE, International League Against Epilepsy.

The 87 MRI scans included 22 cases (25.3%) that were MAP(+). The radiologist detected abnormal findings in 24 cases (27.6%), which were correlated with the FCD (RAD[+]) such as in the presence of cortical thickness or thinning, blurred GM-WM junction, or abnormal signal intensities in T2- or T1-weighted images. Specifically, RAD(+)/MAP(+), RAD(+)/MAP(-), RAD(-)/MAP(+), and RAD(-)/MAP(-) accounted for 21.8%, 17.2%, 3.4%, and 57.5% of the cases, respectively (Table 2). Discordant results between the reports of radiologists and the MAP indicating RAD(+)/MAP(-) and RAD(-)/MAP(+), were observed in 18 patients (20.7%). In the retrospective analysis, the lesions in discordant cases were included in the resection margin completely or partially. All concordant cases showed that the FCD lesions detected by both the MAP and radiologist (MAP[+]/RAD[+]) were completely included in resection margins (Table 3). Among those 18 discordant cases, representative results of 1 of the RAD(-)/MAP(+) cases are shown in Fig. 2, and other analyses with MAP(+) are shown in Supplementary Figs. 1-3 (in the online-only Data Supplement).

Table 2. Detection rates of focal cortical dysplasia by radiologists and MRI postprocessing in each category.

	RAD(+)	RAD(-)
MAP(+)	19 (21.8)	3 (3.4)
MAP(-)	15 (17.2)	50 (57.5)

Data are n (%) values.

MAP, morphometric analysis program.

Table 3. Comparison between the discordant and concordant cases in detected focal cortical dysplasia lesions included in the resection margin.

		Inclusion in resection margin
		Complete	Partial	None
Discordant (n=18)
	MAP(-)/RAD(+) (n=15)	11	4	0
	MAP(+)/RAD(-) (n=3)	2	1	0
Concordant (n=69)
	MAP(+)/RAD(+) (n=19)	19	0	0
	MAP(-)/RAD(-) (n=50)	N/A	N/A	N/A

MAP, morphometric analysis program.

Fig. 2. MRI scans and the postprocessing results of the MAP scans of the patient diagnosed with FCD via pathology. This case was categorized as RAD(-)/MAP(+). A-F: Initially, the radiologist interpreted the T2-weighted (A) and 3D T1-weighted (B) MRI scans as normal. The MAP scans highlighted FCD lesions in the junction (C) the most prominently (marked with the red circle), but also in the extension (D) and thickness (E) maps. Postoperative MRI (F) indicated that a resected epileptogenic focus was correlated with the MAP result. FCD, focal cortical dysplasia; MAP, morphometric analysis program.

Regarding the detection of FCD lesions, the detection rate by the MAP for the test cases of ten types of MRI scanners did not differ according to the two reference scanners (Philips Ingenia CX 3-tesla and Siemens Skyra 3-tesla) except for in only one case. The detection rate was 35.0% when using the same kind of scanner as a reference with that of test case, which means that the test case from Ingenia scanner was analyzed based on the Ingenia reference and Skyra based on the Skyra reference, while it was 22.4% in the analysis when the types of the reference and test case differed. These detection rates did not differ significantly (p=0.26).

The detection rates of ten different scanners are shown in Fig. 3A. All MRI scans of the test cases were performed following the epilepsy protocol in our center (as described above), except in one MRI scan that was performed in March 2020 using the Discovery MR750w scanner (GE Healthcare, https://www.gehealthcare.com) and included an arterial spin labeling perfusion sequence. Among the RAD(-) cases, the MAP detected a lesion in three patients, all of whom had the FCD type IIa pathology, while the most common pathologies (88.2%) with RAD(+)/MAP(-) were FCD types I or III. FCD type I accounted for most of the MAP(-) patients (p<0.001) (Fig. 3B). For RAD(+)/MAP(-) cases, the common radiologic findings were cystic lesion, atrophic lesion, subtle change in signal intensity, and cavernous malformation related to FCD type III, whereas the MAP were specialized in detecting FCD lesions such as a thickened cortex, GM-WM junction, or transmantle sign. Among nine RAD(+)/MAP(-) cases classified as FCD type I, while multiple pathologies in each case were present, radiologists were suspicious of FCD in the reports of five. The radiologic findings in the other four cases were dysembryoplastic neuroepithelial tumor, nonspecific findings (old trauma, infarction, or inflammation), and ganglioglioma.

Fig. 3. Detection of FCD lesion according to MRI scanners and FCD subtypes. A: Different MRI scans of MAP(+) and MAP(-) cases. B: Subtypes of FCD confirmed by pathology after surgery were compared among MAP(+) and MAP(-) cases. ^*Most MAP(-) patients were FCD type I (p<0.001). FCD, focal cortical dysplasia; MAP, morphometric analysis program.

The seizure outcomes after surgery in ILAE class 1 according to detection tools are listed in Table 4. Postoperative seizure outcomes of each case were evaluated in the latest visit to the outpatient center based on the ILAE classification, with class 1 indicating being completely seizure free and no aura.

Table 4. Postoperative seizure outcomes according to detection tools.

	RAD(+)	RAD(-)
MAP(+)	14/19 (73.7)*	0/3 (0)
MAP(-)	8/15 (53.3)	13/50 (26.0)

Data are n (%) values.

^*The number of the cases represented the number of patients who achieved a final seizure-free state.

MAP, morphometric analysis program.

DISCUSSION

This study investigated the feasibility and diagnostic value of the MAP18 as a supportive tool for conventional multimodal and multidisciplinary presurgical evaluations. It was the second largest study on pathology-proven FCD with morphometric analysis. This was the first study that we were aware of to examine postprocessing of MRI data for FCD in Korea aimed at developing a MAP algorithm specialized for a single center using the existing MAP with a database of normal MRI scans from that center.

Each MRI scanner had specific characteristics and there were also fundamental differences in the population, meaning that a well-constructed standard reference of the population and database using different MRI scanners is necessary to enhance the ability to detect FCD lesions. In this study, a normal database that was constructed from MRI scans performed using two different scanners (Philips Ingenia CX 3-tesla and Siemens Skyra 3-tesla) was used as a reference. With the exception of one case, the MAP18 was found to detect true FCD lesions or yield negative results from the test cases independent of the type of machine used, with the standard MRI database as a reference. The MAP therefore still had advantages in detecting FCD lesions in the test case when using a different type of scanner from the reference. The generalizability of processing images from different scanners has been a major issue in imaging software. It was particularly interesting that the MAP for FCD diagnosis might be more applicable with various types of references; for example, a previous pilot study used the MAP and identified FCD in a childhood database using an adult template.¹⁵ However, a well-designed study with a large population is required to obtain ample evidence about the reliable common reference.

All 87 patients in the study underwent presurgical evaluation, including MRI, and they all visited the epilepsy center after surgery at regular intervals. Medical records and the results of laboratory tests, radiologic examinations, and pathologic findings were well documented, so the data analyzed in this study were of high quality. The availability of well-documented data in a center could facilitate the construction of a high-quality database.

Moreover, the MAP could identify FCD type IIa among the RAD(-) cases in this study, while there was also an unusually high number of FCD type I cases (57/87, 65.5%) which was consistent with the other previous study¹⁶ performed at our center involving patients with FCD. A previous study⁸ of 91 patients with FCD found that the MAP was superior to visual analysis, especially in type IIa, which was consistent with our results. However, the detection rate of the MAP combined with visual analysis was as high as 98.0% in that study, which was much higher than the rate of 42.5% observed in our study. One hypothesis to explain the relatively low detection rate in the present study is that our center is a national tertiary hospital, and most of the patients whose diagnoses are difficult to make in local hospitals without definite lesion visit our center. There were therefore many cases of nonlesional epilepsy. Similarly, previous studies involving nonlesional cases also found detection rates that were lower¹⁷,¹⁸,¹⁹,^20,21,²²,²³ than that in the study⁸ of 91 patients with FCD.

The present study also found a relatively low detection rate among nonlesional epilepsy cases of MAP(+) when combined with visual analysis by radiologists, which might be explained by the well-established radiologic reporting system in our center. The radiologists considered clinical information and continued to communicate with clinicians using a multidisciplinary approach and to provide feedback for formal reports. The number of MAP(+) cases that cannot be detected by visual analysis might therefore be relatively small in our center.

The MAP process achieves rapid detection and is relatively simple for clinicians to perform. All steps involved in the pre- and postprocessing of MRI scans took less than 15 minutes for each case. Acquisition of MRI scans via PACS software and all postprocessing and analyses using SPM5 software in MATLAB could be performed at a single workstation, making it efficient for clinicians to use. We therefore suggest that the MAP18 is a promising diagnostic tool to support radiologists, while its benefit could depend on the type of FCD. As a supportive tool, rather than as an alternative to the existing system, the MAP18 could not detect all FCD lesions in the test cases in this study, making them MAP(-) cases. The previously mentioned system of radiologist reporting in our center was also one of the explanations for MAP(-)/RAD(+) cases, and so further studies are necessary to explore the sensitivity and specificity of the MAP18 to determine if it is a better diagnostic tool.

There were several limitations in this study. First, because it was a retrospective study rather than a randomized controlled trial, the patients were not randomly distributed or compared with historical controls adjusted for sex or age. Subtle invisible lesions in the “normal” reference data set could also have affected the final MAP results. In fact, because a 3D T1-weighted sequence was not included in the routine brain MRI protocol in our center, there was an inevitable limitation to constructing the “normal” database using a population with a wide age range. Second, due to the relatively small number of patients, we could not analyze causal or correlational relationships between the detection rate and factors that might have affected the analysis. Third, in some of the MAP(+)/RAD(-) cases, the radiologists might have been biased due to findings of other examinations (electroencephalography or positron-emission tomography). Fourth, the patients in this study had heterogeneous characteristics and might not represent the wider population because the study was performed at a single epilepsy center in Korea. However, using the MAP, clinicians could detect FCD lesions with prominent visualization in addition to radiologic findings, which avoids the underestimations that could happen in presurgical evaluations.

Detecting FCD lesions in patients with drug-resistant epilepsy is very important, especially when they are considered for surgical treatment due to its relationship with surgical outcome. We suggest the MAP as a supportive diagnostic tool for FCD due to its rapid and relative accessibility in detecting FCD lesions. Further prospective controlled studies with larger numbers of patients are warranted to explore its diagnostic value.

Availability of Data and Material

The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.

Supplementary Materials

The online-only Data Supplement is available with this article at https://doi.org/10.3988/jcn.2022.0317.

Supplementary Table 1

MR parameters for each scanner used for post-processing analysis

Supplementary Fig. 1

Among MAP(+) cases, the results of five representative cases are shown. Each result consists of junction, extension, and thickness map along with T1 MRI. Prominently visualized lesion by MAP is indicated with dotted circle (skyblue). MAP, morphometric analysis program.

Supplementary Fig. 2

Among MAP(+) cases, the results of five cases with RAD(+) are shown. Each result consists of junction, extension, and thickness map along with T1 MRI. Prominently visualized lesion by MAP is indicated with dotted circle (skyblue). MAP, morphometric analysis program.

Supplementary Fig. 3

Among MAP(+) cases, the results of four other cases with RAD(+) are shown. Each result consists of junction, extension, and thickness map along with T1 MRI. Prominently visualized lesion by MAP is indicated with dotted circle (skyblue). MAP, morphometric analysis program.