Min Jeong Kwon; Subin Lee; Jieun Park; Sungman Jo; Ji Won Han; Dae Jong Oh; Jun-Young Lee; Joon Hyuk Park; Jae Hyoung Kim; Ki Woong Kim

doi:10.3346/jkms.2023.38.e316

Journal List > J Korean Med Sci > v.38(41) > 1516084259

Go to TopGo to Top Go to BottomGo to Bottom

TOOLS

Kwon, Lee, Park, Jo, Han, Oh, Lee, Park, Kim, and Kim: Textural and Volumetric Changes of the Temporal Lobes in Semantic Variant Primary Progressive Aphasia and Alzheimer’s Disease

Original Article

Neuroscience

Journal of Korean Medical Science 2023; 38(41): e316.

Published online: 18 September 2023

DOI: https://doi.org/10.3346/jkms.2023.38.e316

Textural and Volumetric Changes of the Temporal Lobes in Semantic Variant Primary Progressive Aphasia and Alzheimer’s Disease

Min Jeong Kwon¹

, Subin Lee¹

, Jieun Park¹

, Sungman Jo²

, Ji Won Han^3,⁴

, Dae Jong Oh⁵

, Jun-Young Lee^4,⁶

, Joon Hyuk Park⁷

, Jae Hyoung Kim⁸

, Ki Woong Kim^1,^2,^3,⁴

¹Department of Brain and Cognitive Science, Seoul National University College of Natural Science, Seoul, Korea.

²Department of Health Science and Technology, Graduate School of Convergence Science and Technology, Seoul National University, Seoul, Korea.

³Department of Neuropsychiatry, Seoul National University Bundang Hospital, Seongnam, Korea.

⁴Department of Psychiatry, Seoul National University College of Medicine, Seoul, Korea.

⁵Workplace Mental Health Institute, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, Seoul, Korea.

⁶Department of Neuropsychiatry, Seoul Metropolitan Government-Seoul National University Boramae Medical Center, Seoul, Korea.

⁷Department of Neuropsychiatry, Jeju National University Hospital, Jeju, Korea.

⁸Department of Radiology, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Seongnam, Korea.

Address for Correspondence: Ki Woong Kim, MD, PhD. Department of Neuropsychiatry, Seoul National University College of Medicine, Seoul National University Bundang Hospital, 166 Gumi-ro, Bundang-gu, Seongnam 13620, Republic of Korea. kwkimmd@snu.ac.kr

Received 7 April 2023 Accepted 15 June 2023

The Korean Academy of Medical Sciences

https://creativecommons.org/licenses/by-nc/4.0/

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Background

Texture analysis may capture subtle changes in the gray matter more sensitively than volumetric analysis. We aimed to investigate the patterns of neurodegeneration in semantic variant primary progressive aphasia (svPPA) and Alzheimer’s disease (AD) by comparing the temporal gray matter texture and volume between cognitively normal controls and older adults with svPPA and AD.

Methods

We enrolled all participants from three university hospitals in Korea. We obtained T1-weighted magnetic resonance images and compared the gray matter texture and volume of regions of interest (ROIs) between the groups using analysis of variance with Bonferroni posthoc comparisons. We also developed models for classifying svPPA, AD and control groups using logistic regression analyses, and validated the models using receiver operator characteristics analysis.

Results

Compared to the AD group, the svPPA group showed lower volumes in five ROIs (bilateral temporal poles, and the left inferior, middle, and superior temporal cortices) and higher texture in these five ROIs and two additional ROIs (right inferior temporal and left entorhinal cortices). The performances of both texture- and volume-based models were good and comparable in classifying svPPA from normal cognition (mean area under the curve [AUC] = 0.914 for texture; mean AUC = 0.894 for volume). However, only the texture-based model achieved a good level of performance in classifying svPPA and AD (mean AUC = 0.775 for texture; mean AUC = 0.658 for volume).

Conclusion

Texture may be a useful neuroimaging marker for early detection of svPPA in older adults and its differentiation from AD.

Graphical Abstract

Keywords: Semantic Variant Primary Progressive Aphasia, Alzheimer’s Disease, MRI, Texture, Volume, Temporal Lobe

INTRODUCTION

Semantic variant primary progressive aphasia (svPPA)1 is a rare dementing illness in older adults.2 SvPPA presents with progressive loss of conceptual knowledge with relatively preserved episodic memory,3 whereas Alzheimer’s disease (AD) presents with episodic memory deficits followed by progressive impairment in global cognition.4

Although the typical clinical presentations of svPPA and AD differ, differentiating svPPA from AD is not easy, particularly in the early stage, using only the clinical presentations.5 In structural brain imaging, svPPA typically shows asymmetric atrophic changes in the left temporal pole,6 7 8 entorhinal cortex (ERC),9 or perirhinal cortex.10 Although AD usually presents with atrophic changes in bilateral medial temporal lobes,11 it is more likely to show asymmetrical atrophy in the early stage.12 13 Furthermore, both svPPA and AD show atrophic changes in the amygdala (AMG), hippocampus (HPC), ERC, parahippocampal gyrus (PHG), inferior temporal cortex (ITC), and middle temporal cortex (MTC).7 9 This indicates that svPPA may not be well differentiated from AD using only volumetric measures.

Texture analysis, which allows quantification of the complex interrelationship between contrasts at the individual voxel level, can capture subtle changes in the gray matter more sensitively than volumetric analysis. We previously demonstrated that, in AD, texture changes preceded volume changes in brain magnetic resonance imaging (MRI).14 Therefore, texture analysis may better classify svPPA, AD, and the controls with normal cognition (NC) than volumetric analysis, particularly in the early stage when the neuronal loss is not prominent. In addition, by simultaneously comparing textural and volumetric changes, we may better understand the differences in the early neurodegenerative changes between svPPA and AD than by comparing volume changes alone.

In this study, we compared the textures and volumes of key regions of interest (ROIs) in the temporal lobes between early svPPA, early AD, and NC, and developed models for classifying them.

METHODS

Participants

We enrolled 30 patients with svPPA who visited the dementia clinics of three university hospitals (Seoul National University Bundang Hospital [SNUBH], Seoul Metropolitan Government–Seoul National University Boramae Medical Center [BMC], and Jeju National University Hospital [JNUH]). We enrolled 60 patients with AD from among the visitors to the dementia clinics of SNUBH whose age, sex, education level, and clinical dementia rating (CDR)15 were matched to those of the 30 patients with svPPA. We enrolled 60 controls with NC whose age, sex, and education level were matched to those of the 30 patients with svPPA from the Korean Longitudinal Study on Cognitive Aging and Dementia (KLOSCAD). The KLOSCAD is a nationwide population-based prospective cohort study of older Koreans. In the KLOSCAD, 6,818 community-dwelling Koreans aged ≥ 60 years were randomly sampled from 30 villages and towns across South Korea using residential rosters. The baseline evaluation was conducted in 2010–2012, and follow-up evaluations were conducted every 2 years until 2020.16

Diagnostic evaluation

Geriatric neuropsychiatrists administered standardized diagnostic interviews, including medical histories and physical and neurological examinations, according to the Korean version of the Consortium to Establish a Registry for Alzheimer’s Disease Assessment Packet Clinical Assessment Battery (CERAD-K)17 and the Korean version of the Mini International Neuropsychiatric Interview.18 They also conducted a brain MRI and 18F-Florbetaben positron emission tomography (PET). Research neuropsychologists or trained nurses performed the CERAD-K Neuropsychological Assessment Battery,19 Frontal Assessment Battery,20 Digit Span Test,21 and Geriatric Depression Scale (GDS).22 The CERAD-K Neuropsychological Assessment Battery consists of the verbal fluency test, Boston naming test, Mini-Mental State Examination (MMSE), word list memory test, word list recall test, word list recognition test, constructional praxis test, constructional recall test, and Trail Making Test A/B. We made a payment to SNUPRESS for the CERAD-K Neuropsychological Assessment Battery to clear the copyright.

Subsequently, a panel of geriatric neuropsychiatrists determined the diagnosis and CDR of the participants. They diagnosed svPPA and AD according to the consensus clinical diagnostic criteria for semantic aphasia and associative agnosia proposed by Neary et al.3 and the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association diagnostic criteria for probable AD, respectively.23 All patients with AD were amyloid-β-positive on 18F-Florbetaben PET. Among the 30 patients with svPPA, 18 were amyloid-β-negative on 18F-Florbetaben PET and the rest did not undergo the procedure. The severity of svPPA and AD in all patients was mild (CDR = 0.5 or 1) and matched. All controls with NC were living independently in the community with a CDR of 0 and amyloid-β-negative on 18F-Florbetaben PET.

MR image acquisition and preprocessing

We obtained three-dimensional (3D) structural T1-weighted spoiled gradient echo MR images of the participants within 2 years of their clinical assessments using a 3.0 T Achieva Scanner (Philips Medical Systems; Eindhoven, Netherlands) at the three national university hospitals. The images were acquired using the following parameters: voxel size of 1.0 × 0.5 × 0.5 mm³, 1.0 mm sagittal slice thickness with no inter-slice gap, echo time of 4.6 ms, repetition time of 8.1 ms, flip angle of 8° and a matrix size of 175 × 240 × 240 in the x, y, and z dimensions in SNUBH; voxel size of 1.0 × 1.0 × 1.0 mm³, 1.0 mm sagittal slice thickness with no inter-slice gap, echo time of 4.6 ms, repetition time of 9.9 ms, flip angle of 8° and a matrix size of 180 × 220 × 200 in the x, y, and z dimensions in BMC; voxel size of 1.0 × 1.0 × 1.0 mm³, 1.0 mm sagittal slice thickness with no inter-slice gap, echo time of 3.7 ms, repetition time of 8.2 ms, flip angle of 8° and a matrix size of 190 × 256 × 256 in the x, y, and z dimensions in JNUH.

We used the original Digital Imaging and Communications in Medicine format images and converted them to the Neuroimaging Informatics Technology Initiative format for analysis using MRIcron software. T1 images were bias-corrected to remove intensity inhomogeneity artifacts using the Statistical Parametric Mapping software (version 8, SPM8; Wellcome Trust Centre for Neuroimaging, London, UK; http://www.fil.ion.ucl.ac.uk/spm). We subsequently resliced the bias-corrected T1 images into isotropic voxels (1.0 × 1.0 × 1.0 mm³). Next, we automatically segmented whole brain structures by recon-all streams of FreeSurfer version 6.0 (http://surfer.nmr.mgh.harvard.edu).24 This is a reconstruction process consisting of three steps. In the first stage, motion correction, non-uniform intensity normalization, and skull stripping are conducted. In the second stage, full-scale volumetric labeling is performed with automatic topology fixing. In the final stage, spherical mapping and cortical parcellation are completed. After the recon-all process, we obtained parcellated individual brain masks of nine ROIs, including the AMG, fusiform gyrus (FFG), HPC, PHG, ERC, temporal polar cortex (TPC), ITC, MTC, and superior temporal cortex (STC), where abnormal proteins commonly accumulate in AD and svPPA.6 7 9 25 26 We also obtained paracentral lobule (PaCL), precentral lobule (PrCL), frontal polar cortex (FPC), orbito frontal cortex (OFC), middle frontal cortex (MFC) and superior frontal cortex (SFC) brain masks, which included to the frontal lobe according to the Desikan-Killiany-Tourville atlas in FreeSurfer.27

Analysis of regional volume and texture

We estimated the regional volume of all ROIs using FreeSurfer software.28 We further preprocessed each ROI image prior to texture analysis as follows: We first conducted histogram normalization by removing any outlier voxels with intensity values beyond the range, μ − 3σ, μ + 3σ (where μ is the mean signal intensity of the gray levels and σ is its standard deviation), to guard against partial volume effects.29 We normalized each gray matter voxel with respect to the participant’s mean cerebrospinal fluid signal intensity in the lateral ventricle regions to correct for any interindividual variations in scaling factors. Finally, we performed quantization in each ROI image, rescaling all signal intensity values to a uniform range of 32 to reduce discrete values, thereby avoiding statistical problems related to sparse matrices during the calculation of texture features.30

We subsequently conducted 3D gray-level co-occurrence matrix (GLCM) analysis to extract texture features (from each preprocessed ROI image using MATLAB R2020a [MathWorks, Natick, MA, USA]). The GLCM is an N × N matrix, where N is the total number of gray levels in the image. Each element (i,j) of the matrix reports the frequency of specific pairs of gray-level values, including the reference voxel i and the neighboring voxel j, which occur at distance d and direction θ. We looked at voxel pairs within a distance of d = 1 of each other (directly adjacent voxels) in 13 different directions, resulting in 13 GLCMs per ROI. Each GLCM is normalized so that the frequency of each voxel pair is divided by the total number of voxels in the ROI. Haralick texture features, which are mathematical equations that utilize the average GLCM as input, are subsequently calculated.31 32 33 We compared the autocorrelation in the allo/periallocortical ROIs (HPC, PHG and ERC) and the contrast in neocortical ROIs (ITC, MTC, STC, FFG and TPC) and AMG between the diagnostic groups, according to our previous work.34

Statistical analysis

Continuous variables were compared using one-way analysis of variance (ANOVA) and categorical variables using chi-square tests between the diagnostic groups. In each ROI, we compared the volume and texture features (autocorrelation or contrast) between the diagnostic groups using ANOVA with Bonferroni posthoc comparisons. Furthermore, we calculated regional z-scores for the AD and svPPA groups. The z-score map indicates the magnitude of volume and texture differences in each clinical group compared with the control group.

We developed texture-based and volume-based models for classifying NC, AD and svPPA using logistic regression with a forward selection of variables. A five-fold cross-validation was used to train and validate the models. We estimated the classification performance of the models using receiver operator characteristic (ROC) curve analysis and compared the area under the ROC curve (AUC) between the models according to Hanley and McNeil.35

A two-tailed P value < 0.05 was considered statistically significant in all analyses. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 25.0 (IBM Corporation, Armonk, NY, USA) on Windows and MedCalc for Windows version 18.11.3 (MedCalc Software, Mariakerke, Belgium).

Ethics statement

This study was approved by the Institutional Review Board (IRB) of the Seoul National University Bundang Hospital and the requirement for informed consent was waived (IRB No. B-2005-615-001).

RESULTS

The MMSE and GDS scores of the AD and svPPA groups were comparable but higher than those of the NC group. The estimated total intracranial volumes were comparable among the three diagnostic groups (Table 1).

Table 1

Participant characteristics

Characteristics		NC (n = 60)	AD (n = 60)	svPPA (n = 30)	Statistics^a
Characteristics		NC (n = 60)	AD (n = 60)	svPPA (n = 30)	P value	Posthoc
Age, yr		73.13 ± 5.98	75.05 ± 7.84	71.53 ± 8.26	0.082	-
Women		33 (55.0)	41 (68.3)	15 (50.0)	0.168	-
Education, yr		12.85 ± 4.04	11.83 ± 4.82	12.53 ± 5.41	0.480	-
CDR						-
	0	60 (100)	-	-
	0.5	-	34 (56.7)	17 (56.7)
	1	-	26 (43.3)	13 (43.3)
MMSE, points		28.12 ± 1.84	19.23 ± 4.85	20.33 ± 4.96	< 0.001	a > b, c
GDS, points		6.88 ± 5.50	12.58 ± 7.01	11.60 ± 7.91	< 0.001	a < b, c
eTIV, cc		1,511.21 ± 168.82	1,505.31 ± 133.14	1,542.96 ± 185.31	0.555	-

Values are presented as mean ± standard deviation or number (%).

NC = normal cognition, AD = Alzheimer’s disease, svPPA = semantic variant primary progressive aphasia, CDR = Clinical Dementia Rating, MMSE = Mini Mental State Examination, GDS = Geriatric Depression Scale, eTIV = estimated total intracranial volume.

^aOne-way analysis of variance for continuous variables and chi-square test for categorical variables with Bonferroni posthoc comparisons.

The volumes and textures of all ROIs differed among the three diagnostic groups (Table 2). Compared with the NC group, both the svPPA and AD groups showed smaller volumes and higher texture values in all ROIs in both hemispheres. The differences in volume and texture values between the svPPA and AD groups were more prominent in the left hemisphere. Compared to the AD group, the svPPA group showed a smaller volume in the bilateral TPC and in three ROIs (ITC, MTC, and STC) in the left hemisphere, and a higher texture value in two ROIs (TPC and ITC) bilaterally and three ROIs (ERC, MTC, and STC) of the left hemisphere. There were significant svPPA-AD differences in the texture values of the left ERC and right ITC where the volumes were comparable between the svPPA and AD groups.

Table 2

Comparison of the regional volumes and textures between controls with NC and patients with AD and svPPA

ROIs		Volume, mm³					Texture^b
ROIs		NC	AD	svPPA	P value^a	Posthoc^a	NC	AD	svPPA	P value^a	Posthoc^a
Left hemisphere
	AMG	1,264.73 ± 204.95	982.59 ± 176.53	894.45 ± 241.3	< 0.001	a > b, c	24.18 ± 3.45	25.69 ± 3.02	27.25 ± 4.62	0.001	a < c
	FFG	8,365.6 ± 918.19	7,632.53 ± 1,036.73	7,166.77 ± 1,080.88	< 0.001	a > b, c	21.42 ± 2.1	23.96 ± 2.55	24.68 ± 3.25	< 0.001	a < b, c
	HPC	3,536.36 ± 313.64	2,861.19 ± 378.39	3,001.83 ± 631.41	< 0.001	a > b, c	272.00 ± 9.03	282.58 ± 19.71	289.00 ± 20.50	< 0.001	a < b, c
	PHG	1,722.58 ± 238.45	1,492.6 ± 266.57	1,438.2 ± 339.56	< 0.001	a > b, c	266.37 ± 11.13	275.45 ± 19.06	284.33 ± 25.06	< 0.001	a < b, c
	ERC	1,986.85 ± 328.50	1,465.97 ± 346.65	1,300.33 ± 427.04	< 0.001	a > b, c	278.61 ± 15.99	294.29 ± 22.60	317.38 ± 43.71	< 0.001	a < b < c
	TPC	2,397.17 ± 348.59	2,147.55 ± 413.02	1,684.33 ± 569.96	< 0.001	a > b > c	24.52 ± 2.67	27.28 ± 3.9	34.02 ± 8.69	< 0.001	a < b < c
	ITC	10,275.80 ± 1,503.49	8,909.63 ± 1,419.12	7,840.43 ± 1,812.34	< 0.001	a > b > c	19.18 ± 1.78	21.90 ± 2.12	24.31 ± 2.59	< 0.001	a < b < c
	MTC	9,807.88 ± 1,199.89	8,597.58 ± 1,275.82	7,606.00 ± 1,562.43	< 0.001	a > b > c	19.82 ± 1.98	21.97 ± 1.85	23.99 ± 3.08	< 0.001	a < b < c
	STC	10,738.55 ± 1,417.92	9,630.03 ± 1,083.10	8,907.47 ± 1,317.20	< 0.001	a > b > c	20.86 ± 1.78	23.40 ± 1.91	24.66 ± 2.36	< 0.001	a < b < c
Right hemisphere
	AMG	1,475.75 ± 207.16	1,129.11 ± 204.36	1,084.79 ± 321.79	< 0.001	a > b, c	24.85 ± 2.42	26.42 ± 2.88	26.71 ± 4.68	0.007	a < b, c
	FFG	8,096.22 ± 942.17	7,050.4 ± 1,075.81	6,814.73 ± 1,531.54	< 0.001	a > b, c	22.87 ± 3.16	26.82 ± 2.86	27.17 ± 3.37	< 0.001	a < b, c
	HPC	3,684.07 ± 349.81	2,925.44 ± 411.83	3,168.99 ± 794.38	< 0.001	a > b, c	270.19 ± 9.70	283.26 ± 20.01	290.31 ± 29.91	< 0.001	a < b, c
	PHG	1,669.7 ± 239.37	1,450.2 ± 239.57	1,398.37 ± 276.32	< 0.001	a > b, c	266.55 ± 14.71	280.21 ± 20.93	283.73 ± 26.93	< 0.001	a < b, c
	ERC	1,908.68 ± 355.11	1,354.72 ± 344.78	1,167.00 ± 389.55	< 0.001	a > b, c	274.02 ± 12.65	305.62 ± 33.53	297.29 ± 36.43	< 0.001	a < b, c
	TPC	2,443.63 ± 337.14	2,200.55 ± 380.1	1,836.13 ± 446.89	< 0.001	a > b > c	25.82 ± 3.20	28.36 ± 3.31	32.83 ± 8.98	< 0.001	a < b < c
	ITC	9,593.18 ± 1,606.59	7,956.40 ± 1,372.85	7,430.33 ± 2,107.68	< 0.001	a > b, c	22.25 ± 1.99	25.40 ± 2.00	27.48 ± 5.35	< 0.001	a < b < c
	MTC	10,380.93 ± 1,314.50	8,821.08 ± 1,370.85	8,319.30 ± 2,099.94	< 0.001	a > b, c	20.06 ± 1.82	23.64 ± 2.17	24.80 ± 5.30	< 0.001	a < b, c
	STC	10,187.23 ± 1,218.65	9,021.35 ± 1,029.65	8,660.40 ± 1,866.28	< 0.001	a > b, c	21.03 ± 1.45	23.82 ± 1.92	24.35 ± 4.08	< 0.001	a < b, c

Values are presented as mean ± standard deviation.

ROI = regions of interest, NC = normal cognition, AD = Alzheimer’s disease, svPPA = semantic variant primary progressive aphasia, AMG = amygdala, FFG = fusiform gyrus, HPC = hippocampus, PHG = parahippocampal gyrus, ERC = entorhinal cortex, TPC = temporal polar cortex, ITC = inferior temporal cortex, MTC = middle temporal cortex, STC = superior temporal cortex.

^aOne-way analysis of variance with Bonferroni post hoc comparisons.

^bAutocorrelation for the hippocampus, parahippocampal gyrus, and entorhinal cortex and contrast for other regions of interest.

SvPPA and AD showed different patterns of neurodegenerative changes (Fig. 1). In AD, the atrophy was prominent in HPC and ERC, and texture changes were prominent in bilateral STC, in addition to HPC and ERC, indicating that the progression of neurodegeneration in AD may be medial-to-lateral. However, in svPPA, the atrophy was prominent in TPC and ERC, and texture changes were prominent in bilateral ITC, in addition to TPC and ITC, indicating that the progression of neurodegeneration in svPPA may be anterior-to-posterior.

Fig. 1

Z-score map of regional volume and texture changes in the AD and svPPA groups. (A) Regional volume changes in each clinical group compared to the control group. (B) Regional texture changes in each clinical group compared to the control group.

AD = Alzheimer’s disease, svPPA = semantic variant primary progressive aphasia.

As shown in Table 3, the volumes of the left ERC and MTC were most frequently selected in the volume-based model, whereas the texture values of the left ITC, left ERC, and right HPC were most frequently selected in the texture-based model, for classifying svPPA and NC. The mean AUCs of the texture- and volume-based models for classifying svPPA and NC were comparable (mean = 0.914, standard error of the mean [SEM] = 0.046 for the texture-based model; mean = 0.894, SEM = 0.031 for the volume-based model; P = 0.718) (Fig. 2A).

Table 3

Models for differentiating patients with semantic variant primary progressive aphasia patients from controls with NC and patients with AD

Variables		Volume-based model^a					Texture-based model^a
Variables		Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5
From NC
	Intercept	−9.558	−4.046	−3.148	−3.170	−3.738	−8.321	−5.712	−5.101	−3.670	−6.012
	Left AMG	-	-	-	-	−2.421	-	-	-	-	1.691
	Left FFG	-	-	-	-	-	-	-	-	-	-
	Left HPC	−2.095	-	-	-	-	-	-	-	-	-
	Left PHG	-	-	-	-	-	-	1.275	0.838	0.719	-
	Left ERC	−3.578	−1.460	−1.466	−1.381	−1.359	-	-	-	-	-
	Left TPC	-	−0.732	-	-	-	-	1.440	-	-	-
	Left ITC	-	-	-	-	-	3.342	1.609	2.794	2.072	2.404
	Left MTC	−3.747	−1.112	−0.808	−0.737	-	-	-	-	-	-
	Left STC	-	-	-	-	-	-	-	-	-	2.341
	Right AMG	−1.774	-	-	-	-	2.350	1.515	-	-	-
	Right FFG	-	-	−0.733	−0.913	-	-	-	-	-	-
	Right HPC	-	-	-	-	-	2.709	-	1.167	0.876	1.430
	Right PHG	-	-	-	-	−1.221	-	-	-	-	-
	Right ERC	-	-	-	-	-	-	-	-	-	-
	Right TPC	-	-	-	-	-	-	-	-	-	-
	Right ITC	-	-	-	-	-	-	-	-	-	-
	Right MTC	-	-	-	-	-	-	-	-	-	-
	Right STC	-	-	-	-	-	-	-	-	-	-
	AUC	0.819	0.861	0.986	0.944	0.861	0.750	0.889	0.931	1.000	1.000
From AD
	Intercept	−1.747	−2.810	−1.609	−1.346	−0.269	−1.945	−2.142	−2.083	−2.128	−1.538
	Left AMG	-	-	-	-	-	-	-	-	-	-
	Left FFG	-	-	-	-	-	-	-	-	-	-
	Left HPC	-	-	-	0.595	-	-	-	-	-	-
	Left PHG	-	-	-	-	-	-	-	-	-	-
	Left ERC	-	-	-	-	-	0.414	0.351	0.379	0.333	-
	Left TPC	−0.754	−0.968	−0.695	−0.723	−0.627	0.592	0.741	0.398	0.481	0.421
	Left ITC	-	-	-	-	-	-	-	-	-	-
	Left MTC	-	-	-	-	-	-	-	-	-	-
	Left STC	-	−0.757	-	-	-	-	-	-	-	-
	Right AMG	-	-	-	−0.952	-	-	-	-	-	-
	Right FFG	-	-	-	-	-	-	-	-	-	-
	Right HPC	-	-	-	1.007	0.623	-	-	-	-	-
	Right PHG	-	-	-	-	-	-	-	-	-	-
	Right ERC	-	-	-	-	-	−0.238	−0.243	-	-	-
	Right TPC	-	-	-	-	-	-	-	-	-	-
	Right ITC	-	-	-	-	-	-	-	-	-	-
	Right MTC	-	-	-	-	-	-	-	-	-	-
	Right STC	-	-	-	-	-	-	-	-	-	-
	AUC	0.625	0.597	0.750	0.681	0.639	0.722	0.833	0.667	0.722	0.931

NC = normal cognition, AD = Alzheimer’s disease, AMG = amygdala, FFG = fusiform gyrus, HPC = hippocampus, PHG = parahippocampal gyrus, ERC = entorhinal cortex, TPC = temporal polar cortex, ITC = inferior temporal cortex, MTC = middle temporal cortex, STC = superior temporal cortex, AUC = area under the receiver operating characteristic curve.

^aForward selection in logistic regression analysis with five-fold cross-validation.

Fig. 2

Comparison of the performance of the texture- and volume-based models for differentiating patients with semantic variant primary progressive aphasia from cognitively normal controls and patients with AD. (A) Models for differentiating patients with semantic variant primary progressive aphasia from cognitively normal controls. (B) Models for differentiating patients with semantic variant primary progressive aphasia from those with AD.

AUC = area under the receiver operating characteristic curve, CI = confidence interval, AD = Alzheimer’s disease.

In classifying svPPA and AD, the volume of the left TPC was consistently selected in the volume-based model, whereas the texture values of the left TPC and ERC were most frequently selected in the texture-based model. The mean AUCs of the texture-based model was higher than that of the volume-based model for classifying svPPA and AD (mean = 0.775, SEM = 0.047 for texture-based model; mean = 0.658, SEM = 0.027 for volume-based model; P = 0.031) (Fig. 2B).

Both the AD and svPPA groups showed smaller volumes and higher texture values in several frontal ROIs (FPC, OFC, MFC and SFC) compared with the NC group. However, volumes and texture values were comparable in all frontal ROIs between svPPA and AD (Supplementary Table 1).

DISCUSSION

This study demonstrated that the texture features of the temporal lobe could classify svPPA and NC as accurately as the volume features of the temporal lobe, and svPPA and AD far more accurately than the volume features. To the best of our knowledge, this is the first study to simultaneously compare the volume and texture of the temporal and frontal lobes between NC and patients with AD, and svPPA.

In most previous ROI-based and voxel-based neuroimaging studies on svPPA, atrophy was prominent in or confined to temporal lobes.36 37 38 39 40 In the current study, both the svPPA and AD groups showed smaller volumes of all ROIs in both hemispheres than NC. This indicates that atrophic changes, although not as prominent as in the ROIs in the left hemisphere, may also be present in those in the right hemisphere in the early stage of svPPA. A previous voxel-based morphometric study also showed that patients with svPPA had smaller AMG, anterior HPC, TPC, anterior FFG, MTC, and posterior insula in both hemispheres than NC.41 In another ROI-based study, patients with svPPA showed a smaller TPC in both hemispheres and smaller AMG, FFG, and ITC in the left hemisphere.7

Although svPPA and AD have degenerative changes in the temporal lobes in common, svPPA shows a different pattern of temporal neurodegeneration from that in AD. As demonstrated in Fig. 1, neurodegenerative changes may progress medio-laterally from the HPC in AD and antero-posteriorly from the TPC in svPPA. These results are consistent with those of the previous studies. In AD, atrophy starts in the HPC before the onset of symptoms and progresses slowly to other temporal structures with disease progression. In contrast, in svPPA, cortical atrophy is most prominent in temporal poles, where atrophy does not usually occur in normal aging, and progresses along the neural pathways connected to temporal poles and progresses faster in the left hemisphere.42 43 44 In a previous study that directly compared the volumes of temporal ROIs between patients with svPPA and AD, patients with svPPA showed smaller AMG, FFG, ITC, and MTC in the left hemisphere and smaller TPC in both hemispheres than in patients with AD.7 Although these results were also in line with those of the current study, the volumes of the left AMG and FFG, which were smaller in svPPA than in AD in the previous study, were comparable between the two groups in the current study. This discrepancy between the findings of the current and previous studies may be attributed to the differences in the participants’ severity of dementia between the studies. The mean CDR of the participants in the previous study was 0.9 ± 0.6, whereas in the current study it was 0.7 ± 0.3, indicating that the severity was milder in the current study than in the previous study. The volumes of the left AMG and FFG were more likely to differ between the svPPA and AD groups with advancing dementia severity. Although temporal atrophy occurs in both hemispheres in svPPA, the interhemispheric asymmetry of temporal atrophy may decrease with advancing severity of svPPA until temporal atrophy in the right hemisphere becomes as severe as that in the left.

In our previous studies, we showed that the textures of the HPC, precuneus, and posterior cingulate cortex predicted the conversion from mild cognitive impairment to AD earlier and with higher accuracy than hippocampal volume,14 and that the textures of T1-weighted MRI reflect microstructural changes associated with regional tau burden.34 A higher contrast indicates a heterogeneous image, which means that neighboring voxels have very different intensities, and a higher autocorrelation reflects the overall brightening of regional signal intensities. Microstructural changes associated with neurofibrillary tangles may selectively increase the signal intensities of voxels in specific layers, possibly because signal intensity may become brighter with the increase in white matter-like tissue that replaces normal gray matter tissue. Therefore, the texture of brain MRI may change in regions where microstructural changes occur because of the accumulation of tau or TDP type C proteins.39 45 46 47 In the current study, we clearly demonstrated that patients with svPPA showed higher texture values than NC in all ROIs in both hemispheres, where they showed reduced volumes than NC. In addition, we showed that patients with svPPA had higher texture values than patients with AD in some ROIs, such as the left ERC and right ITC, where regional volumes were comparable. These results suggest that neurodegeneration in these ROIs may begin earlier and/or progress faster in svPPA than in AD.

The accuracy of the logistic regression model for differentiating svPPA from NC using temporal textures was as excellent as that using temporal volumes, indicating that temporal textures may differentiate svPPA from NC as accurately as temporal volumes (Fig. 2A). However, in differentiating svPPA from AD, only the model using temporal textures showed a good accuracy (AUC = 0.775, Fig. 2B). Therefore, NC, AD and svPPA can be better differentiated if the texture features of the temporal lobe are used rather than the volume features.

This study has some limitations. First, this study did not include data on neuropathology. Therefore, this study could not directly show whether the texture differences between the diagnostic groups were attributable to the neuropathologies of svPPA or AD. Second, this study employed a cross-sectional design, and thus could not show whether the texture differences preceded volume differences. Third, the sample size was limited; thus, our classification models need to be validated in future studies.

Notes

Funding: This study was supported by a grant from the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (Grant No. HI09C1379 [A092077]) and a grant of the Korea Dementia Research Project through the Korea Dementia Research Center (KDRC), funded by the Ministry of Health & Welfare and Ministry of Science and ICT, Republic of Korea (Grant No. HU20C0015).

Disclosure: The authors have no potential conflicts of interest to disclose.

Author Contributions:

Conceptualization: Kwon MJ.
Data curation: Park J, Oh DJ, Lee JY, Park JH, Kim JH.
Formal analysis: Jo S.
Project administration: Kim KW.
Resources: Lee S.
Software: Lee S.
Supervision: Han JW, Kim KW.
Writing - original draft: Kwon MJ.
Writing - review & editing: Kwon MJ, Han JW, Kim KW.

References

1. Gorno-Tempini ML, Hillis AE, Weintraub S, Kertesz A, Mendez M, Cappa SF, et al. Classification of primary progressive aphasia and its variants. Neurology. 2011; 76(11):1006–1014. PMID: 21325651.

2. Ratnavalli E, Brayne C, Dawson K, Hodges JR. The prevalence of frontotemporal dementia. Neurology. 2002; 58(11):1615–1621. PMID: 12058088.

3. Neary D, Snowden JS, Gustafson L, Passant U, Stuss D, Black S, et al. Frontotemporal lobar degeneration: a consensus on clinical diagnostic criteria. Neurology. 1998; 51(6):1546–1554. PMID: 9855500.

4. Greene JD, Baddeley AD, Hodges JR. Analysis of the episodic memory deficit in early Alzheimer’s disease: evidence from the doors and people test. Neuropsychologia. 1996; 34(6):537–551. PMID: 8736567.

5. Liscic RM, Storandt M, Cairns NJ, Morris JC. Clinical and psychometric distinction of frontotemporal and Alzheimer dementias. Arch Neurol. 2007; 64(4):535–540. PMID: 17420315.

6. Davies RR, Graham KS, Xuereb JH, Williams GB, Hodges JR. The human perirhinal cortex and semantic memory. Eur J Neurosci. 2004; 20(9):2441–2446. PMID: 15525284.

7. Galton CJ, Patterson K, Graham K, Lambon-Ralph MA, Williams G, Antoun N, et al. Differing patterns of temporal atrophy in Alzheimer’s disease and semantic dementia. Neurology. 2001; 57(2):216–225. PMID: 11468305.

8. Mummery CJ, Patterson K, Price CJ, Ashburner J, Frackowiak RS, Hodges JR. A voxel-based morphometry study of semantic dementia: relationship between temporal lobe atrophy and semantic memory. Ann Neurol. 2000; 47(1):36–45. PMID: 10632099.

9. Chan D, Fox NC, Scahill RI, Crum WR, Whitwell JL, Leschziner G, et al. Patterns of temporal lobe atrophy in semantic dementia and Alzheimer’s disease. Ann Neurol. 2001; 49(4):433–442. PMID: 11310620.

10. La Joie R, Landeau B, Perrotin A, Bejanin A, Egret S, Pélerin A, et al. Intrinsic connectivity identifies the hippocampus as a main crossroad between Alzheimer’s and semantic dementia-targeted networks. Neuron. 2014; 81(6):1417–1428. PMID: 24656258.

11. Pereira JB, Cavallin L, Spulber G, Aguilar C, Mecocci P, Vellas B, et al. Influence of age, disease onset and ApoE4 on visual medial temporal lobe atrophy cut-offs. J Intern Med. 2014; 275(3):317–330. PMID: 24118559.

12. Derflinger S, Sorg C, Gaser C, Myers N, Arsic M, Kurz A, et al. Grey-matter atrophy in Alzheimer’s disease is asymmetric but not lateralized. J Alzheimers Dis. 2011; 25(2):347–357. PMID: 21422522.

13. Wu X, Wu Y, Geng Z, Zhou S, Wei L, Ji GJ, et al. Asymmetric Differences in the gray matter volume and functional connections of the amygdala are associated with clinical manifestations of Alzheimer’s disease. Front Neurosci. 2020; 14:602. PMID: 32670008.

14. Lee S, Lee H, Kim KW. Alzheimer’s Disease Neuroimaging Initiative. Magnetic resonance imaging texture predicts progression to dementia due to Alzheimer disease earlier than hippocampal volume. J Psychiatry Neurosci. 2020; 45(1):7–14. PMID: 31228173.

15. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology. 1993; 43(11):2412–2414.

16. Han JW, Kim TH, Kwak KP, Kim K, Kim BJ, Kim SG, et al. Overview of the Korean longitudinal study on cognitive aging and dementia. Psychiatry Investig. 2018; 15(8):767–774.

17. Lee JH, Lee KU, Lee DY, Kim KW, Jhoo JH, Kim JH, et al. Development of the Korean version of the Consortium to Establish a Registry for Alzheimer’s Disease Assessment Packet (CERAD-K): clinical and neuropsychological assessment batteries. J Gerontol B Psychol Sci Soc Sci. 2002; 57(1):47–53.

18. Yoo SW, Kim YS, Noh JS, Oh KS, Kim CH, Namkoong K, et al. Validity of Korean version of the mini-international neuropsychiatric interview. Anxiety Mood. 2006; 2(1):50–55.

19. Lee DY, Lee KU, Lee JH, Kim KW, Jhoo JH, Kim SY, et al. A normative study of the CERAD neuropsychological assessment battery in the Korean elderly. J Int Neuropsychol Soc. 2004; 10(1):72–81. PMID: 14751009.

20. Kim TH, Huh Y, Choe JY, Jeong JW, Park JH, Lee SB, et al. Korean version of frontal assessment battery: psychometric properties and normative data. Dement Geriatr Cogn Disord. 2010; 29(4):363–370. PMID: 20424455.

21. Wechsler D. Wechsler Memory Scale-Revised. San Antonio, TX, USA: Psychological Corporation;1987.

22. Kim JY, Park JH, Lee JJ, Huh Y, Lee SB, Han SK, et al. Standardization of the Korean version of the geriatric depression scale: reliability, validity, and factor structure. Psychiatry Investig. 2008; 5(4):232–238.

23. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology. 1984; 34(7):939–944. PMID: 6610841.

24. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron. 2002; 33(3):341–355. PMID: 11832223.

25. Hodges JR, Patterson K. Semantic dementia: a unique clinicopathological syndrome. Lancet Neurol. 2007; 6(11):1004–1014. PMID: 17945154.

26. Landqvist Waldö M, Frizell Santillo A, Passant U, Zetterberg H, Rosengren L, Nilsson C, et al. Cerebrospinal fluid neurofilament light chain protein levels in subtypes of frontotemporal dementia. BMC Neurol. 2013; 13(1):54. PMID: 23718879.

27. Klein A, Tourville J. 101 labeled brain images and a consistent human cortical labeling protocol. Front Neurosci. 2012; 6:171. PMID: 23227001.

28. Fischl B. FreeSurfer. Neuroimage. 2012; 62(2):774–781. PMID: 22248573.

29. Collewet G, Strzelecki M, Mariette F. Influence of MRI acquisition protocols and image intensity normalization methods on texture classification. Magn Reson Imaging. 2004; 22(1):81–91. PMID: 14972397.

30. Patel MB, Rodriguez JJ, Gmitro AF. Effect of gray-level re-quantization on co-occurrence based texture analysis. In : 2008 15th IEEE International Conference on Image Processing; October 12-15, 2008; San Diego, CA, USA. Piscataway, NJ, USA: IEEE;p. 585–588.

31. Puetz AM, Olsen R. Haralick Texture Features Expanded Into the Spectral Domain. Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XII. Bellingham, WA: USA International Society for Optics and Photonics;2006. p. 623311.

32. Haralick RM, Shanmugam K, Dinstein IH. Textural features for image classification. IEEE Trans Syst Man Cybern. 1973; SMC-3(6):610–621.

33. Ortiz A, Palacio AA, Górriz JM, Ramírez J, Salas-González D. Segmentation of brain MRI using SOM-FCM-based method and 3D statistical descriptors. Comput Math Methods Med. 2013; 2013:638563. PMID: 23762192.

34. Lee S, Kim KW. Alzheimer’s Disease Neuroimaging Initiative. Associations between texture of T1-weighted magnetic resonance imaging and radiographic pathologies in Alzheimer’s disease. Eur J Neurol. 2021; 28(3):735–744. PMID: 33098172.

35. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983; 148(3):839–843. PMID: 6878708.

36. Graham KS, Simons JS, Pratt KH, Patterson K, Hodges JR. Insights from semantic dementia on the relationship between episodic and semantic memory. Neuropsychologia. 2000; 38(3):313–324. PMID: 10678697.

37. Kim EJ, Ku BD, Na DL. Semantic dementia. Dement Neurocogn Disord. 2005; 4(1):10–13.

38. Jun BS, Park JH. Frontotemporal dementia. Korean J Biol Psychiatry. 2016; 23(3):69–79.

39. Landin-Romero R, Tan R, Hodges JR, Kumfor F. An update on semantic dementia: genetics, imaging, and pathology. Alzheimers Res Ther. 2016; 8(1):52. PMID: 27915998.

40. Yang J, Pan P, Song W, Shang HF. Quantitative meta-analysis of gray matter abnormalities in semantic dementia. J Alzheimers Dis. 2012; 31(4):827–833. PMID: 22699847.

41. Gorno-Tempini ML, Dronkers NF, Rankin KP, Ogar JM, Phengrasamy L, Rosen HJ, et al. Cognition and anatomy in three variants of primary progressive aphasia. Ann Neurol. 2004; 55(3):335–346. PMID: 14991811.

42. Scahill RI, Schott JM, Stevens JM, Rossor MN, Fox NC. Mapping the evolution of regional atrophy in Alzheimer’s disease: unbiased analysis of fluid-registered serial MRI. Proc Natl Acad Sci U S A. 2002; 99(7):4703–4707. PMID: 11930016.

43. Collins JA, Montal V, Hochberg D, Quimby M, Mandelli ML, Makris N, et al. Focal temporal pole atrophy and network degeneration in semantic variant primary progressive aphasia. Brain. 2017; 140(2):457–471. PMID: 28040670.

44. Rogalski E, Cobia D, Martersteck A, Rademaker A, Wieneke C, Weintraub S, et al. Asymmetry of cortical decline in subtypes of primary progressive aphasia. Neurology. 2014; 83(13):1184–1191. PMID: 25165386.

45. Hodges JR, Mitchell J, Dawson K, Spillantini MG, Xuereb JH, McMonagle P, et al. Semantic dementia: demography, familial factors and survival in a consecutive series of 100 cases. Brain. 2010; 133(Pt 1):300–306. PMID: 19805492.

46. Josephs KA, Hodges JR, Snowden JS, Mackenzie IR, Neumann M, Mann DM, et al. Neuropathological background of phenotypical variability in frontotemporal dementia. Acta Neuropathol. 2011; 122(2):137–153. PMID: 21614463.

47. Rohrer JD, Lashley T, Schott JM, Warren JE, Mead S, Isaacs AM, et al. Clinical and neuroanatomical signatures of tissue pathology in frontotemporal lobar degeneration. Brain. 2011; 134(Pt 9):2565–2581. PMID: 21908872.

SUPPLEMENTARY MATERIAL

Supplementary Table 1

Comparison of the regional volumes and textures in the frontal lobe between controls with NC and patients with AD and svPPA

jkms-38-e316-s001.doc

TOOLS

Similar articles