Comparison of manual tracing versus a semiautomatic radial measurement method in temporal lobe MRI volumetry for pharmacoresistant epilepsy

Neuroradiology (2007) 49:189–201 DOI 10.1007/s00234-006-0171-3

DIAGNOSTIC NEURORADIOLOGY

Comparison of manual tracing versus a semiautomatic radial measurement method in temporal lobe MRI volumetry for pharmacoresistant epilepsy Christian-Andreas Mueller & Jasmin Scorzin & Roy Koenig & Horst Urbach & Rolf Fimmers & Josef Zentner & Thomas-Nicolas Lehmann & Johannes Schramm

Received: 3 May 2006 / Accepted: 4 October 2006 / Published online: 28 November 2006 # Springer-Verlag 2006

Abstract Introduction The aim of this study was to test a modified radial semiautomated volumetry technique (radial divider technique, RDT) versus the manual volumetry technique (MVT) for proportionality of temporal subvolumes in 30 patients with drug-resistant temporal lobe epilepsy. Methods Included in the study were 30 patients (15 female, 15 male; mean age 39.6 years) with pharmacoresistant epilepsy (mean duration 26.6 years). MRI studies were performed preoperatively on a 1.5-T scanner. All image processing steps and volume measurements were performed using ANALYZE software. The volumes of six subregions were measured bilaterally; these included the superior C.-A. Mueller and J. Scorzin contributed equally to this work. C.-A. Mueller (*) : J. Scorzin : J. Schramm Department of Neurosurgery, University of Bonn, Sigmund-Freud Str. 25, 53105 Bonn, Germany e-mail: [email protected] R. Koenig : H. Urbach Department of Radiology Neuroradiology, University of Bonn, Bonn, Germany R. Fimmers Institute of Medical Biometry, Informatics and Epidemiology, University of Bonn, Bonn, Germany J. Zentner Department of Neurosurgery, University of Freiburg, Freiburg, Germany T.-N. Lehmann Department of Neurosurgery, Charité-University Medicine Berlin, Berlin, Germany

temporal gyrus (STG), middle+inferior temporal gyrus (MITG), fusiform gyrus (FG), parahippocampal gyrus (PHG), amygdala (AM), and hippocampus (HP). Linear regression was used to investigate the relationship between the comparable subvolumes obtained with MVT and RDT. Results Very high correlations (R2 >0.95) between RDT and MVT were observed for the STG+MITG and the STG+ MITG+FG, but low correlations for the PHG subvolumes and the combined PHG + HP +AM subvolumes. These observations were independent of the side of the pathology and of hemisphere. Conclusion The two measurement techniques provided highly reliable proportional results. This series in a homogeneous group of TLE patients suggests that the much quicker RDT is suitable for determining the volume of temporolateral and laterobasal temporal lobe compartments, of both the affected and the non-affected side and the right and left hemisphere. Keywords MRI . Volumetry . Pharmacoresistant epilepsy . Temporal lobe epilepsy Introduction Around 20–30% of patients with temporal lobe epilepsy (TLE) do not become seizure-free despite optimal drug therapy. As seizures nearly always interfere with the quality of life, there is no acceptable seizure frequency [22] and surgical resection of the epileptogenic area is warranted [65]. In surgery for TLE, more tailored types of resection have attracted increasing attention. According to several authors, the extent of lateral or mesial resection can influence seizure outcome as well as neuropsychological

190

outcome [12, 34, 36, 37, 48, 50, 52, 59, 63, 64]. Only a few studies, including our own, have been devoted to the relationship between the type or extent of resection and postoperative seizure outcome [8, 9, 39, 46, 64] or between the extent of resection and neuropsychological sequelae [11, 12, 23, 24, 28, 29, 64]. MR imaging has facilitated attempts to obtain accurate, in vivo volumetric measurements of the whole brain and its substructures. These measurements are increasingly being considered as a diagnostic measure in epilepsy (for an extensive review see Duncan [16]). Most studies have used manual tracing of temporal substructures to determine volume changes [9, 30–32, 34, 57]; however, this method requires extensive rater training, is time-consuming, and has the potential risk of rater bias. Techniques for timeconserving and presumably more reliable automatic segmentation of temporal substructures have been reported [1, 2, 26, 27]. Although these algorithms reduce to a large extent dependency on raters, semiautomated tracing may be more variable than manual procedures, for example because of computational procedures. The aim of this study was to test a modified radial semiautomated volumetry (RDT) versus the manual volumetry technique (MVT) for proportionality of temporal subvolumes in 30 patients with drug-resistant temporal lobe epilepsy.

Neuroradiology (2007) 49:189–201

consent, after being fully informed, to the necessary examinations and to their participation in this study. MRI acquisition All the patients were studied with a 1.5-T scanner (Gyroscan ACS-NT; Philips Medical Systems, Best, The Netherlands), with an epilepsy-dedicated MRI protocol that has been reported on previously [56]. A part of this protocol is a 3-D T1-weighted gradient-echo sequence. It has a slice thickness of 1 mm, a field of view (FOV) of 256 mm, a rectangular field of view (RFOV) of 84%, a repetition time (TR) of 15.3 ms, an echo time (TE) of 3.59 ms, a number of signal averages (NSA) of 1, 140 slices and a 30° flip angle resulting in isotropic 1 mm3 voxels. To align coronal slices perpendicular to the long axis of the hippocampus, a reference line parallel to the temporal horn of the lateral ventricle was defined in the midsagittal pilot (chiasmaticocommisural line, CH-PC) and coronal slices were obtained perpendicular to this line. This method is widely accepted by many researchers to reduce volume averaging and to obtain superior visualization with high efficiency and reliability in volumetric measurement [4, 9, 35, 38, 42, 49]. Nevertheless, scan postprocessing reformatting procedures are necessary to achieve an exact and reliable adjustment in all orientations. MRI analysis and postprocessing

Materials and methods Subjects The subjects who volunteered for this study were patients with pharmacoresistant epilepsy; they comprised of 16 females and 15 males. One individual was excluded due to poor scan quality (temporobasal scan artifacts), resulting in a final sample group of 30 patients. The exclusion criteria for acceptability of MRI data were temporobasal and temporomesial scan artifacts, motion artifacts caused by lack of cooperation by the patient, or missing sequences as qualitatively determined by two experienced neuroradiologists (H.U. and R.K.) and the tracer (C.A.M.). All the subjects had been diagnosed and presurgically evaluated according to previously described algorithms [11, 12, 40, 65]. Their mean age was 39.6±12.7 years, ranging from 17.9 to 61.5 years. The mean duration of epilepsy was 26.6±15.1 years , ranging from 7 to 57 years. The most frequent neuropathological diagnosis was hippocampal sclerosis (HS, n=22), followed by ganglioglioma (n=4), focal cortical dysplasia (n=1), dysembryoplastic neuroepithelial tumor (n=1). In two patients, no neuropathological alterations were found. The local ethical committee approved the study. All the subjects gave their written

3-D T1-weighted MRI datasets were transferred via a local network to a PC and converted into an ANALYZEcompatible format (DICOM files). The ANALYZE software system is entirely built upon a toolkit of optimized functions that are organized into a software development library called AVW. The AVW imaging library is a collection of over 600 functions that are accessible to software developers for building advanced image-based application solutions. Analyze is an integration of the full functionality represented in the AVW toolkit with an intuitive Windows-based interface. All image-processing steps and volume measurements were performed using ANALYZE PC AVW 3.1 and 4.0 software package (Biomedical Imaging Resource, Mayo Foundation, Rochester, Minn.). The trained rater (C.A.M.), who was blind to the side of seizure, performed the image processing. Moreover, the time between measuring the MVT and the RDT was more than 1 year; so exact knowledge or memory of the measured volumes in MVT after this period was unlikely. The datasets underwent automated intensity normalization, spatial filtering and correction for field nonhomogeneities, due to radiofrequency nonuniformity, since these may benefit quantitative measurements [53]. Aligning the coronal slices perpendicular to the hippocam-

Neuroradiology (2007) 49:189–201

pus length axis was achieved through a 3-D registration tool. The CH-PC line is defined as tangential to the superior border of the optical chasm (CH) anteriorly and to the inferior border of the posterior commissure (PC) posteriorly. It is situated at the midbrain diencephalic junction in the midsagittal slice. The close parallelism of the CH-PC to the direction of the parallel sulcus, the lateral fissure, and the plane of the temporal horns of the lateral ventricles, and roughly to the choroidal fissure, is clearly seen, and therefore it seems to be the orientation of choice for horizontal cuts, particularly in the investigation of the temporal lobes (TLs) [54]. Coronal sections orthogonal to the CH-PC will be perpendicular or nearly perpendicular to the hippocampal length axis and are extremely suitable for volumetric measurements in this region of interest. Manual volumetry technique Volume measurements of six subregions were performed bilaterally, i.e. 12 subregions in each patient: the amygdala (AM), hippocampus (HP), parahippocampal gyrus (PHC), fusiform gyrus (FG), combined inferior and medial temporal gyrus (MITG) and superior temporal gyrus (STG). The anatomical landmarks and boundaries that were used follow basically the descriptions of Duvernoy [17–19] and Pantel et al. [49]. Manual tracing began with the generation of auxiliary guideline traces on sagittal cuts, starting with one of the lateral slices. They were necessary for providing a neuroanatomically correct separation of the regions of interest, especially of the border between amygdalae and the hippocampal head, and for obtaining the complete extent of the temporal gyri and the hippocampal tail more easily. The procedure followed the method described by Pantel et al. [49]. Most of the efferent pathways of the hippocampus, including the fimbria and the fornix, were excluded from measurement. These structures are normally easy to recognize on the MRI, due to the fact that they contain myelinated fibers and appear as white matter (WM). Measuring isolated hippocampal gray matter rather than assessing combined white and gray matter is considered to be able to achieve a higher sensitivity and accuracy in detecting volume differences [51] since gray matter lesions mainly constitute classical hippocampal sclerosis, for example [18]. Auxiliary lines set on the sagittal orientation serve as a tracing guideline in the coronal slices afterwards. The region of interest is finally defined on coronal slices since coronal slices oriented perpendicularly to the length axis of the structure of interest are said to achieve better efficiency and reliability in volumetric measurement [4, 35, 42, 45, 49, 51]. At the most anterior portion of the hippocampal head, the tracing starts at the slice, which for the first time displays at least guiding lines defining an area.

191

While tracing the hippocampal head, the temporal horn of the lateral ventricle serves as a lateral border. In addition, the hippocampal head is delineated medially by the endorhinal sulcus. Dorsally, the uncal recess sometimes demarcates it. The ventral border of the hippocampal head laterally is demarcated by the WM of the TL. Medially, the ventral border is delimited by following a line defined by the subiculum– WM interface. Laterally, the temporal stem and the lateral ventricle demarcate the hippocampal body; dorsally, the signal of the cerebrospinal fluid (CSF) of the lateral ventricle, and again the alveus is defined as a border. The CSF of the ambient cistern serves as medial boundary, together with the visible crus cerebri. On the ventral aspect, again the excluded WM of the TL delimits the hippocampus. Laterally, the hippocampus tail is delineated by the ascending crus of the fornix and by the CSF of the atrium of the lateral ventricle. Dorsally, the border is defined by the pulvinar of the thalamus. The dorsal limit is geared to the auxiliary lines in the sagittal view. On the terminal slices, the lower border of the splenium of the corpus callosum marks the dorsal limit. Medially, the CSF of the quadrigeminal cistern and, more caudally, the fasciola cinerea and the gyrus fasciolaris as well as the gyrus of Andreas Retzius become visible, which are excluded from the measurement. Ventrally, the WM of the TL (parahippocampal gyrus) is the limit. On anterior slices, the inferior border of the AM is marked by WM, alveus and/or CSF, and separates the AM from the PHC. The medial superior border is the endorhinal sulcus. On T1-weighted images, the GM signal of the inferior portions of the putamen and claustrum, which come close to the AM, and the delicate areas of WM surrounding them produce sufficient contrast to provide a visually appreciable delineation of the superior extent. The AM is demarcated at the posterior portion by the optic tract. On posterior slices, the inferior border of the AM is marked by the appearance of the HP. The WM tract (temporal stem) forms the lateral border. The medial border is marked by the medial margins of the TL, which borders the CSF. Concerning the remaining regions of interest, we followed in most parts the description of Moran et al. [46]. With regard to the PHC, the superior boundary follows the inferior and inferomedial surface of the HP or AM. It continues medially along the superior surface of the subicular region, along the inferomedial surface of the AM, and follows the medial surface of the TL before turning laterally to the medial bank of the collateral sulcus. At the nadir of this sulcus, the boundary is extended as a straight line to the underlying GM–WM matter interface. From here, a straight line is extended to the lateral most point of the temporal horn, or as a line perpendicular to the inferior surface of the HP or AM, respectively. The floor of the temporal horn is followed along until the hippocampal or amygdala trace is met, thus

192

Neuroradiology (2007) 49:189–201

closing the trace. The PHC arbitrarily terminates anteriorly at the slice in which the anterior commissure (AC) is seen. The substructures of the STG, MITG, and FG were delineated according to the protocol of Moran et al. [46]. To ensure the definitive adhesion of the STG, MITG, FG, and PHG, the posterior boundary is marked by the anterior calcarine sulcus. The regions were automatically calculated by pixel counting, and for each subregion as well (for visualization, see Fig. 1b). Radial divider technique For radial measurement, we used the radial divider tool of ANALYZE 3.1 and 4.0 software, which enables the radial division of a selected region. It is easy to handle and learn. The tool allows (1) a starting angle to be determined (this specifies the angle at which the radial divider is rotated around its center), and (2) an autocenter (this automatically computes the center location by taking an average of the x and y coordinates of all the pixels in the region) (Fig. 2). Using the same datasets, TLE subvolumes were again measured using the RDT. The manually traced contours of the STG, MITG, FG, and PHG in the coronal orientation were fused in the ROI module on both sides (1→2 icon, Fig. 2). The icon was put on autocenter in order to minimize subjective measurement influences. Taking the coronary cuts from the pole posteriorly, one radial division line was placed on the first coronal cut, thus subdividing the TL into two sectors (the superior sector, I, roughly corresponding to the STG in the MVT; the inferior sector, II, to the MITG in the MVT; Fig. 1a). For the first radial on the right side, we chose an angle of 145°; on the left side 45°, measured against the horizontal line (Fig. 1a). We divided the TL into four sectors (two radial division lines)—similar to the partition described by Awad et al. [3]—as soon as we identified the cut of the WM tract of the temporal stem. When we identified this slice we measured also the FG and PHG. In contrast to the procedure described by Awad et al., the four sectors with their spatial orientation were kept right to the posterior border, the calcarine sulcus. The angle of the radial on the right side was 45° from this slice posteriorly to the calcarine sulcus, on the left side 35° (Fig. 1c, Fig. 2). We chose the angles in consideration of the anatomical characteristics in these areas. Thus the upper quadrant (I) of the RDT roughly corresponded to the STG in the MVT, the lateral quadrant (II) to the MITG, the lower quadrant (III) to the FG, and the medial quadrant (IV) to the mesial substructures (PHG, AM and HP; Fig. 1b,c). While measuring the different subvolumes, it was noticeable that the MVT required about 8 hours, the RDT 2 hours. However, a complete measurement with the RDT, without the fusion of the manually traced substructures, required about 4 hours. Although measuring speed

Fig. 1 Boundaries of preoperative temporal structures on MRI. Selected slices from a 47-year-old female patient with HS. a The TL substructures were delineated by RDT anteriorly from the temporal stem. One radial division line was placed on the first coronal cut; thus subdividing the TL in two sectors, on the right side with an angle α= 145°, on the left side α=45° measured against the horizontal line. Sector I corresponding roughly to the STG in the MVT, sector II to the MITG in the MVT. b Preoperative coronal slices at a more posterior level, through the temporal stem, with the temporal substructures being delineated by the MVT. c Temporal substructures were delineated by RDT on the same slice as in b. The division leads to four sectors/quadrants. The angle of the radial was from this slice posterior to the calcarine sulcus on the right side α=45°, on the left side α=35°. Quadrant I roughly corresponds to the STG, II to the MITG, III to the FG, and IV to the PHG, AM, and HP in the MVT (STG superior temporal gyrus, MITG middle and inferior temporal gyri, FG fusiform gyrus, PHG parahippocampal gyrus, AM amygdala, HP hippocampus)

Neuroradiology (2007) 49:189–201

193

Fig. 2 Screenshot of the ROI menu, with the ANALYZE software package adjustment of the radial divider submenu, and a typical measurement condition in a posterior position through the temporal stem

depended also on the experience of the investigator, there was an obvious time-saving. Intrarater reliability Intrarater reliability was calculated with the intraclass correlation coefficient (ICC; ICC=(variance between the individuals−variance between first and second measurement)/variance between the individuals). The ICC for repeated tracings in ten participants (MVT) was satisfactory, e.g. for left FG 0.93, right FG 0.91; or for left HP 0.92, right HP 0.90; and left AM 0.98, right AM 0.93.

Statistical analysis Linear regression was used to investigate the relationship between the subvolumes measured using the MVT and those measured using the RDT. Corrected R2 was used to assess the degree of linear relationship. The two methods were regarded as practically proportional if the corrected R2 was above 0.95 and the intercept of the linear regression was not found to be different from zero. To achieve a more detailed statement about the two measurement methods, we tested with the Bland-Altman-Plot (Fig. 3a,b) [6, 7].

194

Neuroradiology (2007) 49:189–201

Results

Fig. 3 Bland-Altman plots for temporolateral subvolumes in 30 patients. For abbreviations, see legend to Fig. 2. a Classical absolute difference (Bland-Altman) plot for mean value between the RDT (I+II+III quadrant) and MVT (STG+MITG+FG) left side. b BlandAltman difference plot for percentage between RDT (I+ II+III quadrant) and MVT (STG+MITG+FG) left side

Regression analysis was used to compare the temporolateral compartments, i.e. MVT (STG+MITG) vs. RDT (quadrant I+II), and MVT (STG+MITG+FG) vs. RDT (quadrant I+II+III), and the temporomesial compartments, i.e. MVT (PHG) vs. RDT (quadrant IV), MVT (HP) vs. RDT (quadrant IV), and MVT (PHG+AM+HP) vs. RDT (quadrant IV) (Fig. 1a–c). The analysis was performed using SPSS 12.

In detail, for the right STG+MITG, the corrected R2 was 0.967 (Fig. 2a); for the left side, 0.972 (Fig. 2b). Moreover, we observed a corrected R2 of 0.988 for the STG, MITG and FG on the right side as a combined subvolume (Fig. 2c). On the left side, there was a corrected R2 of 0.983 (Fig. 2d). A smaller correlation at the temporomesial substructure of PHG on both sides was found (right side corrected R2 0.674, Fig. 2e; left side corrected R2 0.475, data not shown). A slight correlation was observed regarding the HP volume alone—a structure of particular interest in temporal epilepsy—on both sides (right side corrected R2 0.192, Fig. 2f; left side corrected R2 0.044, data not shown). A similar observation was made when the semiautomated PHG volume and the manually-traced combined PHG, HP and AM were tested. On the right side, the corrected R2 was 0.629 (Fig. 2g), and on the left side 0.318 (data not shown). Individually, the corrected R2 for the healthy side for STG+ MITG was 0.954 (Fig. 4a) and for the affected side, 0.97 (Fig. 4b). The STG+MITG+FG on the non-affected side (corrected R2 0.953; Fig. 4c) was similarly well correlated with the affected side (corrected R2 0.98; Fig. 4d). As assumed, the corrected R2 of the PHG on the healthy side was 0.233 (Fig. 4e), compared to that on the affected side (corrected R2 0.55, data not shown). As expected, the corrected R2 of HP on the affected side was 0.218 (Fig. 4f), and on the healthy side the corrected R2 was 0.03 (data not shown). Regression of PHG (RDT) versus PHG + HP + AM (MVT) averaged on the healthy side revealed a corrected R2 =0.189 (Fig. 4g), and on the pathological side, a corrected R2 of 0.58 (data not shown). The regression analysis showed that the temporolateral substructure volumes (STG+MITG, STG +MITG+FG) were proportional comparing the two methods. In the classical Bland-Altman plot, we observed that the difference in the measured value growths was linear between RDT and MVT (range 9969.55–3226.9 mm3). Therefore we calculated in a second step the relative difference in percentage (range 24.95–6.30%). No obvious relationship between the relative difference between the methods and the extent of the measured values could be seen (for example, see for all patients STG+MITG+FG left side; Fig. 5a,b). Seven main results were observed: (1) there was a very high correlation between the semiautomated RDT and MVT for the STG+MITG, (2) there was a very high correlation between the two techniques for the combined subvolumes STG+MITG+FG, (3) there was a low correlation between the two techniques for the PHG subvolume, (4) there was a low correlation between the two techniques

Neuroradiology (2007) 49:189–201

195

Fig. 4 Relative variation (%) between the two measurement techniques for the volume of temporal substructures in 30 patients, using the semiautomated RDT and the MVT. a, b Correlation between the two measurement techniques for the right (a) and left (b) lateral neocortical combined subvolumes STG+MITG. c, d Correlation between the two measurement techniques for the right (c) and left (d) lateral plus basolateral neocortical combined subvolumes STG+

MITG+FG. e–g Correlation between the two measurement techniques for the right side temporomesial subvolumes PHG (e), HP (f) and combined PHG+AM+HP (g) (STG superior temporal gyrus, MITG middle and inferior temporal gyri, FG fusiform gyrus, PHG parahippocampal gyrus, HP hippocampus; RDT quadrant I ≈ STG of the MVT, RDT quadrant II ≈ MITG of the MVT, RDT quadrant III ≈ FG of the MVT, RDT quadrant IV ≈ PHG/AM/HP of the MVT)

for the HP subvolume, (5) there was a low correlation between the two techniques for the combined PHG+HP+ AM subvolumes, (6) the two measurement techniques provided proportional results because the intercept was not different from zero (combined volumes of STG+MITG, and STG+MITG+FG), and ((7) these observations were independent both in terms of pathology and of hemisphere.

In summary, our results in a homogeneous group of TLE patients suggest that the much quicker RDT is suitable for determining the volume of lateral compartments, such as STG+MITG, or STG+MITG+FG combined, both for the affected and the non-affected sides and the right and left hemispheres. For the purpose of exploring relationships between TL volumes and other parameters, such as seizure

196

Neuroradiology (2007) 49:189–201

Fig. 4 (continued)

duration, extent of resection, freedom from seizure, or neuropsychology, the RDT may save significant time in obtaining volumetric data.

Discussion Our results showed that the MVT and the semiautomated RDT gave equivalent results for the volume of the TL for the temporobasal and laterobasal compartments but not for the temporomesial sectors. However, the aim of this study

was not to show that the two methods led to the same results but rather to what extent are volume measurements by the MVT and RDT proportional or equivalent to each other. Future experiments, e.g. phantom measurements or normal control groups, are necessary to determine which method is better for the correct determination of true volumes. Volumetric 3D MRI produces images with excellent anatomical definition. The technique of hippocampal volumetry has been established for more than a decade [14, 33] to explore TLE and other neurological diseases

Neuroradiology (2007) 49:189–201

197

Fig. 5 Relative variation (%) between the two measurement techniques for the volume of temporal subvolumes in 30 subjects. a, b Correlation between the two measurement techniques for the healthy (a) and pathological (b) sides for the combined lateral subvolumes STG+MITG. c, d Correlation between the two measurement tech-

niques for the healthy (c) and pathological (d) sides for the combined subvolumes STG+MITG+FG. e–g Correlation between the two measurement techniques for mesial temporal subvolumes PHG (healthy side) (e), HP (affected side) (f), and combined PHG+AM+ HP (healthy side) (g). For abbreviations, see legend to Fig. 4

such as Alzheimer’s disease and Parkinson’s disease [41, 44]. In particular, relevant contributions have been made by several groups studying the causes and effects of patholomorphology in chronic epilepsy using TL volumetry [5, 10, 15, 34, 58, 60, 61]. Furthermore, TL volumetry may be used to explore the relationships between extent of resection of TL substructures and outcome in regard to seizures and neuropsychology [8, 9, 39, 46, 62]. According to these results, the time needed to complete

the volumetric task may be considerably reduced. Several authors have followed similar approaches using semiautomated methods [21, 25, 43, 47, 55]. Moreover, some groups have developed fully automated segmentation and labeling techniques [13, 20]. They also labeled small regions, e.g. the nucleus caudatus [13] and the hippocampus [20, 44], with good correlations between the segmentation techniques, but they compared semiautomated methods with fully automated segmentation meth-

198

Neuroradiology (2007) 49:189–201

Fig. 5 (continued)

ods with different angulation methods. However, as reports on RDT are few, it has remained an open question as to whether the RDT technology provides reliable results in all compartments. Therefore no meaningful discussion of other results from the application of the RDT feature to TL volumetry is possible. In our study, the temporomesial substructures (HP, AM, and PHG) did not correlate between RDT with MVT, which corresponds with our expectation. There are several possible reasons for this: (1) the pathology affects mainly

the mesial lobe, leading to a larger variability in volumes, resulting in a larger variability of tracing; (2) the basically unavoidable irregularity in measurement values is larger in the corresponding mesial volumes than in the lateral subvolumes; and (3) the RDT tends towards including non-relevant temporomesial structures, e.g. when we calculated correlations for the HP or the PHG (MVT vs. RDT). In the case of the RDT, there was volume included that was definitely not associated with the HP or PHG measured in the MVT. However, we also found low

Neuroradiology (2007) 49:189–201

correlations for the combined PHG+HP+AM subvolumes between RDT and MVT. In this case, the RDT did not tend towards over-including or under-including relevant structures significantly (see Fig. 1a,b). Additional RDT studies are necessary to resolve the problem, e.g. the ability to choose a different starting angle in the consideration of the anatomical characteristics in these areas. Furthermore, the RDT offers the advantage that the surface delineation of the substructures is easier because the arbitrarily drawn demarcation lines of STG, MITG and FG are not confounding factors with the RDT. The values resulting from the semiautomated RDT for the mesial sectors, i.e. AM, HP, and PHG, do not fit well enough with the values obtained with the MVT. For the purpose outlined at the beginning—saving time in TL volumetry—without sacrificing reliability, the RDT may therefore replace manual volumetry of the STG, MITG, and FG subcompartments of the TL, except for those studies where each subvolume needs to be known separately. The method is not appropriate for the determination of absolute volumes of the TL subcompartments are examined, but is well-suited if relative volume changes are to be assessed, e.g. as in this study comparing pre- and postoperative volumes. The lower correlation of the mesial subvolumes does not invalidate the RDT approach for those studies where a detailed and precise volumetry of each simple subvolume from the mesial group (HP, AM, or PHG) is not a prerequisite. Acknowledgements This project was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG) as part of the TR3 Collaborative Research Project (Sonderforschungsbereich) “Mesial Temporal Lobe Epilepsy”. We thank Prof. Elger and colleagues of the Department of Epileptology, University of Bonn, for evaluating patients and for their close and longstanding collaboration. We also thank Petra Suessmann, Sabine Richter and Sandra Thulke for technical support and statistical analysis.

Conflict of interest statement of interest.

We declare that we have no conflict

References 1. Ashton EA, Berg MJ, Parker KJ, Weisberg J, Chen CW, Ketonen L (1995) Segmentation and feature extraction techniques, with applications to MRI head studies. Magn Reson Med 33:670–677 2. Ashton EA, Parker KJ, Berg MJ, Chen CW (1997) A novel volumetric feature extraction technique with applications to MR images. IEEE Trans Med Imaging 16:365–371 3. Awad IA, Katz A, Hahn JF, Kong AK, Ahl J, Luders H (1989) Extent of resection in temporal lobectomy for epilepsy. I. Interobserver analysis and correlation with seizure outcome. Epilepsia 30:756–762

199 4. Bartzokis G, Altshuler LL, Greider T, Curran J, Keen B, Dixon WJ (1998) Reliability of medial temporal lobe volume measurements using reformatted 3D images. Psychiatry Res 82:11–24 5. Bernasconi N, Bernasconi A, Caramanos Z, Dubeau F, Richardson J, Andermann F, Arnold DL (2001) Entorhinal cortex atrophy in epilepsy patients exhibiting normal hippocampal volumes. Neurology 56:1335–1339 6. Bland JM, Altman DG (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307–310 7. Bland JM, Altman DG (1995) Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet 346:1085–1087 8. Burneo JG, Bilir E, Faught E, Morawetz R, Knowlton RC, Martin R, Kuzniecky RI (2003) Significance of fornix atrophy in temporal lobe epilepsy surgery outcome. Arch Neurol 60:1238–1242 9. Cendes F, Andermann F, Gloor P, Evans A, Jones-Gotman M, Watson C, Melanson D, Olivier A, Peters T, Lopes-Cendes I (1993) MRI volumetric measurement of amygdala and hippocampus in temporal lobe epilepsy. Neurology 43:719–725 10. Cendes F, Caramanos Z, Andermann F, Dubeau F, Arnold DL (1997) Proton magnetic resonance spectroscopic imaging and magnetic resonance imaging volumetry in the lateralization of temporal lobe epilepsy: a series of 100 patients. Ann Neurol 42:737–746 11. Clusmann H, Schramm J, Kral T, Helmstaedter C, Ostertun B, Fimmers R, Haun D, Elger CE (2002) Prognostic factors and outcome after different types of resection for temporal lobe epilepsy. J Neurosurg 97:1131–1141 12. Clusmann H, Kral T, Fackeldey E, Blumcke I, Helmstaedter C, von Oertzen J, Urbach H, Schramm J (2004) Lesional mesial temporal lobe epilepsy and limited resections: prognostic factors and outcome. J Neurol Neurosurg Psychiatry 75:1589–1596 13. Collins DL, Holmes CJ, Peters TM, Evans AC (1995) Automatic 3-D model-based Neuroanatomical segmentation. Hum Brain Mapp 3:190–208 14. Cook MJ, Fish DR, Shorvon SD, Straughan K, Stevens JM (1992) Hippocampal volumetric and morphometric studies in frontal and temporal lobe epilepsy. Brain 115:1001–1015 15. Duncan JS, Bartlett P, Barker GJ (1996) Technique for measuring hippocampal T2 relaxation time. AJNR Am J Neuroradiol 17:1805–1810 16. Duncan JS (2002) MRI studies. Do seizures damage the brain? Prog Brain Res 135:253–261 17. Duvernoy CS (1988) The human hippocampus. An atlas of applied anatomy. J.F. Bergmann, Munich 18. Duvernoy H (1998) The human hippocampus: functional anatomy, vascularization and serial sections with MRI. Springer, Berlin Heidelberg 19. Duvernoy H (1999) The human brain. Surface, three-dimensional sectional anatomy with MRI, and blood supply. Springer, Wien, New York 20. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, van der Kouwe A, Killiany R, Kennedy D, Klaveness S, Montillo A, Makris N, Rosen B, Dale AM (2002) Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 33:341–355 21. Freeborough PA, Fox NC, Kitney RI (1997) Interactive algorithms for the segmentation and quantitation of 3-D MRI brain scans. Comput Methods Programs Biomed 53:15–25 22. Gilliam F (2001) Health outcomes in persons with epilepsy. Neurol Clin 19:465–475 23. Gleissner U, Sassen R, Lendt M, Clusmann H, Elger CE, Helmstaedter C (2002) Pre- and postoperative verbal memory in pediatric patients with temporal lobe epilepsy. Epilepsy Res 51:287–296

200 24. Goldstein LH, Polkey CE (1993) Short-term cognitive changes after unilateral temporal lobectomy or unilateral amygdalo-hippocampectomy for the relief of temporal lobe epilepsy. J Neurol Neurosurg Psychiatry 56:135–140 25. Hagemann G, Lemieux L, Free SL, Krakow K, Everitt AD, Kendall BE, Stevens JM, Shorvon SD (2002) Cerebellar volumes in newly diagnosed and chronic epilepsy. J Neurol 249:1651– 1658 26. Haller JW, Christensen GE, Joshi SC, Newcomer JW, Miller MI, Csernansky JG, Vannier MW (1996) Hippocampal MR imaging morphometry by means of general pattern matching. Radiology 199:787–791 27. Haller JW, Banerjee A, Christensen GE, Gado M, Joshi S, Miller MI, Sheline Y, Vannier MW, Csernansky JG (1997) Threedimensional hippocampal MR morphometry with high-dimensional transformation of a neuroanatomic atlas. Radiology 202:504–510 28. Helmstaedter C, Elger CE (1996) Cognitive consequences of twothirds anterior temporal lobectomy on verbal memory in 144 patients: a three-month follow-up study. Epilepsia 37:171–180 29. Helmstaedter C, Grunwald T, Lehnertz K, Gleissner U, Elger CE (1997) Differential involvement of left temporolateral and temporomesial structures in verbal declarative learning and memory: evidence from temporal lobe epilepsy. Brain Cogn 35:110–131 30. Jack CR Jr, Gehring DG, Sharbrough FW, Felmlee JP, Forbes G, Hench VS, Zinsmeister AR (1988) Temporal lobe volume measurement from MR images: accuracy and left-right asymmetry in normal persons. J Comput Assist Tomogr 12:21–29 31. Jack CR Jr, Twomey CK, Zinsmeister AR, Sharbrough FW, Petersen RC, Cascino GD (1989) Anterior temporal lobes and hippocampal formations: normative volumetric measurements from MR images in young adults. Radiology 172:549–554 32. Jack CR Jr, Bentley MD, Twomey CK, Zinsmeister AR (1990) MR imaging-based volume measurements of the hippocampal formation and anterior temporal lobe: validation studies. Radiology 176:205–209 33. Jack CR Jr, Sharbrough FW, Twomey CK, Cascino GD, Hirschorn KA, Marsh WR, Zinsmeister AR, Scheithauer B (1990) Temporal lobe seizures: lateralization with MR volume measurements of the hippocampal formation. Radiology 175:423–429 34. Jack CR Jr, Sharbrough FW, Cascino GD, Hirschorn KA, O’Brien PC, Marsh WR (1992) Magnetic resonance image-based hippocampal volumetry: correlation with outcome after temporal lobectomy. Ann Neurol 31:138–146 35. Jack CR Jr, Theodore WH, Cook M, McCarthy G (1995) MRIbased hippocampal volumetrics: data acquisition, normal ranges, and optimal protocol. Magn Reson Imaging 13:1057–1064 36. Kanner AM, Kaydanova Y, de Toledo-Morrell L, Morrell F, Smith MC, Bergen D, Pierre-Louis SJ, Ristanovic R (1995) Tailored anterior temporal lobectomy. Relation between extent of resection of mesial structures and postsurgical seizure outcome. Arch Neurol 52:173–178 37. Katz A, Awad IA, Kong AK, Chelune GJ, Naugle RI, Wyllie E, Beauchamp G, Luders H (1989) Extent of resection in temporal lobectomy for epilepsy. II. Memory changes and neurologic complications. Epilepsia 30:763–771 38. Kim JH, Tien RD, Felsberg GJ, Osumi AK, Lee N, Friedman AH (1995) Fast spin-echo MR in hippocampal sclerosis: correlation with pathology and surgery. AJNR Am J Neuroradiol 16:627–636 39. Kitchen ND, Cook MJ, Shorvon SD, Fish DR, Thomas DG (1994) Image guided audit of surgery for temporal lobe epilepsy. J Neurol Neurosurg Psychiatry 57:1221–1227 40. Kral T, Clusmann H, Urbach J, Schramm J, Elger CE, Kurthen M, Grunwald T (2002) Preoperative evaluation for epilepsy surgery (Bonn algorithm). Zentralbl Neurochir 63:106–110

Neuroradiology (2007) 49:189–201 41. Laakso MP, Partanen K, Riekkinen P, Lehtovirta M, Helkala EL, Hallikainen M, Hanninen T, Vainio P, Soininen H (1996) Hippocampal volumes in Alzheimer’s disease, Parkinson’s disease with and without dementia, and in vascular dementia: an MRI study. Neurology 46:678–681 42. Laakso MP, Juottonen K, Partanen K, Vainio P, Soininen H (1997) MRI volumetry of the hippocampus: the effect of slice thickness on volume formation. Magn Reson Imaging 15:263–265 43. Luft AR, Skalej M, Welte D, Kolb R, Burk K, Schulz JB, Klockgether T, Voigt K (1998) A new semiautomated, threedimensional technique allowing precise quantification of total and regional cerebellar volume using MRI. Magn Reson Med 40:143– 151 44. Lye TC, Grayson DA, Creasey H, Piguet O, Bennett HP, Ridley LJ, Kril JJ, Broe GA (2006) Predicting memory performance in normal ageing using different measures of hippocampal size. Neuroradiology 48:90–99 45. Mayhew TM (1992) A review of recent advances in stereology for quantifying neural structure. J Neurocytol 21:313–328 46. Moran NF, Lemieux L, Maudgil D, Kitchen ND, Fish DR, Shorvon SD (1999) Analysis of temporal lobe resections in MR images. Epilepsia 40:1077–1084 47. Mori E, Yoneda Y, Yamashita H, Hirono N, Ikeda M, Yamadori A (1997) Medial temporal structures relate to memory impairment in Alzheimer’s disease: an MRI volumetric study. J Neurol Neurosurg Psychiatry 63:214–221 48. Nayel MH, Awad IA, Luders H (1991) Extent of mesiobasal resection determines outcome after temporal lobectomy for intractable complex partial seizures. Neurosurgery 29:55–60; discussion 60–61 49. Pantel J, O’Leary DS, Cretsinger K, Bockholt HJ, Keefe H, Magnotta VA, Andreasen NC (2000) A new method for the in vivo volumetric measurement of the human hippocampus with high neuroanatomical accuracy. Hippocampus 10:752–758 50. Pauli E, Pickel S, Schulemann H, Buchfelder M, Stefan H (1999) Neuropsychologic findings depending on the type of the resection in temporal lobe epilepsy. Adv Neurol 81:371–377 51. Sheline YI, Wang PW, Gado MH, Csernansky JG, Vannier MW (1996) Hippocampal atrophy in recurrent major depression. Proc Natl Acad Sci USA 93:3908–3913 52. Siegel AM, Wieser HG, Wichmann W, Yasargil GM (1990) Relationships between MR-imaged total amount of tissue removed, resection scores of specific mediobasal limbic subcompartments and clinical outcome following selective amygdalohippocampectomy. Epilepsy Res 6:56–65 53. Sled JG, Zijdenbos AP, Evans AC (1998) A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging 17:87–97 54. Tamraz JC, Comair YG (2000) Atlas of regional anatomy of the brain using MRI. Springer, Berlin Heidelberg New York, pp 11–50 55. Tisserand DJ, Pruessner JC, Sanz Arigita EJ, van Boxtel MP, Evans AC, Jolles J, Uylings HB (2002) Regional frontal cortical volumes decrease differentially in aging: an MRI study to compare volumetric approaches and voxel-based morphometry. Neuroimage 17:657–669 56. Urbach H, Hattingen J, von Oertzen J, Luyken C, Clusmann H, Kral T, Kurthen M, Schramm J, Blumcke I, Schild HH (2004) MR imaging in the presurgical workup of patients with drug-resistant epilepsy. AJNR Am J Neuroradiol 25:919–926 57. Urbach H, Siebenhaar G, Koenig R, von Oertzen J, Scorzin J, Kurthen M, Schild HH (2005) Limbic system abnormalities associated with Ammon’s horn sclerosis do not alter seizure outcome after amygdalohippocampectomy. Epilepsia 46:549–555 58. Van Paesschen W, Connelly A, Johnson CL, Duncan JS (1996) The amygdala and intractable temporal lobe epilepsy: a quantitative magnetic resonance imaging study. Neurology 47:1021–1031

Neuroradiology (2007) 49:189–201 59. Walton NH, Goodsman C, McCarter R, Sandeman DR, Bird JM (1999) An analysis of neuropsychological change scores following selective temporal resection of the non-dominant temporal lobe. Seizure 8:241–245 60. Watson C, Andermann F, Gloor P, Jones-Gotman M, Peters T, Evans A, Olivier A, Melanson D, Leroux G (1992) Anatomic basis of amygdaloid and hippocampal volume measurement by magnetic resonance imaging. Neurology 42:1743–1750 61. Watson C, Jack CR Jr, Cendes F (1997) Volumetric magnetic resonance imaging. Clinical applications and contributions to the understanding of temporal lobe epilepsy. Arch Neurol 54:1521– 1531 62. Wendel JD, Trenerry MR, Xu YC, Sencakova D, Cascino GD, Britton JW, Lagerlund TD, Shin C, So EL, Sharbrough FW, Jack CR (2001) The relationship between quantitative T2 relaxometry

201 and memory in nonlesional temporal lobe epilepsy. Epilepsia 42:863–868 63. Wolf RL, Ivnik RJ, Hirschorn KA, Sharbrough FW, Cascino GD, Marsh WR (1993) Neurocognitive efficiency following left temporal lobectomy: standard versus limited resection. J Neurosurg 79:76–83 64. Wyler AR, Hermann BP, Somes G (1995) Extent of medial temporal resection on outcome from anterior temporal lobectomy: a randomized prospective study. Neurosurgery 37:982–990; discussion 990–991 65. Zentner J, Hufnagel A, Wolf HK, Ostertun B, Behrens E, Campos MG, Solymosi L, Elger CE, Wiestler OD, Schramm J (1995) Surgical treatment of temporal lobe epilepsy: clinical, radiological, and histopathological findings in 178 patients. J Neurol Neurosurg Psychiatry 58:666–673