. 2017 Apr 4;124(8):915–964. doi: 10.1007/s00702-017-1717-8

Table 3.

Diagnostic accuracy of the quantitative assessment of regional cerebral atrophy including MR-planimetry, -volumetry, and automated methods for quantitative MRI analysis for the diagnosis of APD

References	Cohort size	Magnetic field	Main results	Discriminator	Accuracy, %
Schulz et al. (1999)	MSA-P 12/MSA-C 17/PD 11/PSP 6/HC 16	1.5 T	↓ Mean striatal and brainstem volumes in MSA-P, MSA-C, and PSP ↓ Cerebellar volume in MSA-C and MSA-P Volumes were TIV-corrected Patients with PD could not be separated from HC and patients with MSA-P could not be separated from patients with PSP	Discriminant function including volumes of brainstem, caudate nucleus, putamen, and cerebellum (stepwise linear discrimination)^a	Overall correct classification 65% 91% of APD and 89% of non-APD classified correctly 75% of PD, 36% of HC, 67% of MSA-P, 76% of MSA-C and 50% of PSP classified correctly
Schulz et al. (1999)	MSA-P 12/MSA-C 17/PD 11/PSP 6/HC 16	1.5 T			Se 76 (MSA-C) Sp 100 (vs. PD), 82 (vs. HC), 100 (vs. PSP)
Cordato et al. (2002)	PSP 21/PD 17/HC 23	1.5 T	↑ Ventricular volume, ↓ whole brain and frontal GM volumes in PSP vs. PD and HC	Frontal GM volume	Se 95 (PSP) Sp 91 (vs. other groups)
Groschel et al. (2004)	PSP 33/CBS 18/HC 22 (including 8 PSP and 7 CBD with a post-mortem confirmed diagnosis)		↓ Brainstem volume (>midbrain) in PSP vs. CBS and HC ↓ Parietal and occipital lobes volumes (>white matter) in CBS vs. PSP and HC ↓ Area of CC in CBS vs. PSP and HC Volumes were TIV-corrected	Discriminant function including midbrain, parietal WM, temporal GM, brainstem, frontal WM, and pons volumes	Overall correct classification 84% 95% of HC, 76% of CBS, 83% of PSP classified correctly
Paviour et al. (2005)	PSP 19/MSA 10/PD 12/HC 12	1.5 T	↓ SCP volume in PSP vs. MSA, PD, and HC Volumes were TIV-corrected	SCP volume	Se 74 (PSP) Sp 77 (vs. other groups)
Paviour et al. (2006b)	MSA-P 9/PSP 18/PD 9/HC 18	1.5 T	↓ Midbrain and SCP volumes in PSP compared to MSA-P, PD and HC ↓ Frontal volume in PSP compared to PD and HC, but not to MSA-P Significant greater third ventricle in PSP compared to HC, but not to MSA-P and PD ↓ Cerebellar, pontine, and posterior inferior volumes in MSA-P compared to PD and HC ↓ Midbrain volume in MSA-P compared to HC No significant differences between PD and HC	Midbrain, SCP, frontal, third ventricle, and whole brain volumes	Se 89 (PSP)^b Sp 97 (vs. MSA-P, PD, and HC)
Paviour et al. (2006b)	MSA-P 9/PSP 18/PD 9/HC 18	1.5 T		SCP, midbrain, pons, and cerebellar volumes	Se 94 (PSP)^b Sp 89 (vs. MSA-P)
Cosottini et al. (2007)	PSP 15/MSA-P 7/HC 14	1.5 T	↓ m_d in PSP vs. HC, but is not significantly different between PSP and MSA-p patients ↓ m_a in PSP vs. MSA-P and HC ↓ Midbrain volume in PSP patients vs. HC, but does not differ with respect to MSA-P ↓ p_a in MSA-P vs. PSP and HC ↓ m_a/p_a ratio is in PSP vs. MSA-P and HC ↑ m_a/p_a ratio in MSA-P vs. HC	m_d	Se 60 (PSP vs. MSA-P and HC) Sp 95 (vs. MSA-P and HC) AUC = 0.81 Se 60 (PSP vs. MSA-P) Sp 86 (PSP vs. MSA-P) AUC = 0.62 (PSP vs. MSA-P)
				m_a	Se 100 (PSP vs. MSA-P and HC) Sp 90 (PSP vs. MSA-P and HC) AUC = 0.99 Se 87 (PSP vs. MSA-P) Sp 100 (PSP vs. MSA-P) AUC = 0.97
				p_a	Se 73 (PSP vs. MSA-P and HC) Sp 62 (PSP vs. MSA-P and HC) AUC = 0.54 Se 100 (PSP vs. MSA-P) Sp 86 (PSP vs. MSA-P) AUC = 0.96
				m_a/p_a ratio	Se 87 (PSP vs. MSA-P and HC) Sp 100 (PSP vs. MSA-P and HC) AUC = 0.96 Se 100 (PSP vs. MSA-P) Sp 100 (PSP vs. MSA-P) AUC = 1.00
				Midbrain volume	Se 87 (PSP vs. MSA-P and HC) Sp 76 (PSP vs. MSA-P and HC) AUC = 0.83 Se 87 (PSP vs. MSA-P) Sp 71 (PSP vs. MSA-P) AUC = 0.77
Lee et al. (2013b)	PD 29/PSP 13/MSA-P 15/HC 21	3.0 T	↓ Volume of caudate, putamen, globus pallidus, and thalamus in PSP and MSA-P vs. PD and HC ↓ Volume of globus pallidus in PSP vs. MSA-P ↓ Volume of putamen in MSA-P vs. PSP ↑ R2* in the putamen in MSA-P vs. PD and HC ↑ R2* in globus pallidus and caudate in PSP vs. PD and HC ↑ R2* in caudate nucleus in PSP vs. MSA-P	Putaminal volume	AUC = 0.83 (MSA-P vs. PD and PSP)
Lee et al. (2013b)	PD 29/PSP 13/MSA-P 15/HC 21	3.0 T		Volume of globus pallidus	AUC = 0.86 (PSP vs. PD and MSA-P)
Baudrexel et al. (2014)	PD 13/PSP 8/MSA-P 11/HC 6	3.0 T	↓ Putaminal volume in MSA-P vs. PD, PSP, and HC	Putaminal volume	Se 54.5 (MSA-P) Sp 100 (vs. other groups) AUC = 0.84
Zanigni et al. (2016)	PSP-RS 23/PD 42	1.5 T	↑ Sagittal MCP_d/coronal SCP_d ratio, p_a/m_aratio, and MRPI in PSP-RS vs. PD ↓ Sagittal MCP_d, coronal SCP_d, p_a, and m_a in PSP-RS vs. PD ↑ DTI MD in SCP, thalamus, putamen, globus pallidus, parieto-occipital WM, pre-frontal WM, right brain emisphere, left brain hemisphere, posterior fossa, brainstem, and in cerebellar hemispheres in PSP-RS vs. PD ↓ DTI FA in SCP, midbrain, parieto-occipital WM, pre-frontal WM, right brain hemisphere, left brain hemisphere, posterior fossa, and in brainstem in PSP-RS vs. PD ↓ Volume of brainstem, globus pallidus, putamen and thalamus in PSP-RS vs. PD ↑ Volume of lateral ventricles in PSP-RS vs. PD	m_a	Se 96 (PSP-RS) Sp 98 AUC = 0.99
				p_a/m_a-ratio	Se 96 (PSP-RS) Sp 90 AUC = 0.97
				MRPI	Se 87 (PSP-RS) Sp 93 AUC = 0.95
				Coronal SCP_d	Se 81 (PSP-RS) Sp 74 AUC = 0.84
				Pons area	Se 87 (PSP-RS) Sp 74 AUC = 0.82
				MD SCP	Se 70 (PSP-RS) Sp 98 AUC = 0.88
				MD pre-frontal WM	Se 90 (PSP-RS) Sp 69 AUC = 0.86
				MD thalamus	Se 70 (PSP-RS) Sp 86 AUC = 0.84
				MD putamen	Se 80 (PSP-RS) Sp 71 AUC = 0.82
				MD posterior fossa	Se 80 (PSP-RS) Sp 83 AUC = 0.90
				MD cerebellar hemispheres	Se 80 (PSP-RS) Sp 74 AUC = 0.86
				MD right brain hemisphere	Se 75 (PSP-RS) Sp 90 AUC = 0.88
				MD left brain hemisphere	Se 65 (PSP-RS) Sp 95 AUC = 0.87
				MD brainstem	Se 85 (PSP-RS) Sp 67 AUC = 0.80
				FA SCP	Se 75 (PSP-RS) Sp 80 AUC = 0.82
				FA parieto-occipital WM	Se 75 (PSP-RS) Sp 80 AUC = 0.82
				FA right brain hemisphere	Se 75 (PSP-RS) Sp 90 AUC = 0.88
				FA left brain hemisphere	Se 65 (PSP-RS) Sp 95 AUC = 0.87
				FA posterior fossa	Se 83 (PSP-RS) Sp 75 AUC = 0.80
				Volume thalamus	Se 73 (PSP-RS) Sp 90 AUC = 0.83
				Volume putamen	Se 93 (PSP-RS) Sp 70 AUC = 0.83
				Volume globus pallidus	Se 93 (PSP-RS) Sp 60 AUC = 0.81
Oba et al. (2005)	PSP 21/MSA-P 25/PD 23/HC 31	1.5 T	↓ m_a in PSP vs. PD and MSA-P ↓ p_a in MSA-p vs. PD and PSP ↓ In m_a/p_a-ratio in PSP vs. PD and MSA-P	m_a	Se 100 (PSP) Sp 91 (PSP vs. non-PSP)
Oba et al. (2005)	PSP 21/MSA-P 25/PD 23/HC 31	1.5 T		m_a/p_a-ratio	Se 100 (PSP) Sp 100 (PSP vs. non-PSP)
Gama et al. (2010)	PSP 20/MSA-P 8/MSA-C 11/PD 21/	1.5 T	↓ m_a in PSP and MSA-P vs. PD, in PSP vs. MSA-C and MSA-P, and in MSA-P vs. MSA-C ↓ p_a in PSP, MSA-P, and MSA-C vs. PD, and in MSA-C and MSA-P vs. PSP ↓ MCP_d in PSP, MSA-C, and MSA-P vs. PD, and in MSA-C vs. PSP ↓ SCP_d in PSP and MSA-C vs. PD, and in PSP vs. MSA-C and MSA-P	m_a	Se 95 (PSP)^b Sp 97 (PSP vs. other groups)
Gama et al. (2010)	PSP 20/MSA-P 8/MSA-C 11/PD 21/	1.5 T		SCP_d	Se 80 (PSP) Sp 100 (vs. other groups)
Quattrone et al. (2008)	MSA-P 19/PD 108/PSP 33/HC 50	1.5 T	↓ m_a and SCP_d in PSP compared to PD, MSA-P, and HC with some overlap of values between groups ↑ p_a/m_a and MCP_d/SCP_d ratios in PSP compared to PD, MSA-P, and HC with some overlap of values between groups ↑ MRPI in PSP compared to PD, MSA-P, and HC without any overlaps values among groups	p_a/m_a ratio	Se 95 (PSP)^b Sp 97 (vs. PD, MSA-P, and HC)
				MCP_d/SCP_dratio	Se 90 (PSP)^b Sp 94 (vs. PD, MSA-P, and HC)
				MRPI	Se 100 (PSP)^b Sp 100 (vs. PD, MSA-P, and HC)
Hussl et al. (2010)	MSA-P 26/PSP 22/PD 75	1.5 T	↓ m_a/p_a-ratio and ↑ MRPI values in PSP compared to MSA-P, PD, and HC	m_a/p_a ratio	Se 64 (PSP)^b Sp 92 (vs. non-PSP), 95 (vs. PD), 85 (vs. MSA-P)
Hussl et al. (2010)	MSA-P 26/PSP 22/PD 75	1.5 T		MRPI	Se 82 (PSP)^b Sp 80 (vs. non-PSP), 76 (vs. PD), 92 (vs. MSA-P)
Longoni et al. (2011)	PSP-RS 10/ PSP-P 10/ PD 25/HC 24	1.5 T	↑ p_a/m_a ratio in PSP-RS and PSP-P vs. PD ↑ MRPI in PSP-RS and PSP-P vs. PD	p_a/m_a ratio	Se 90 (PSP-RS vs. PD) Sp 96 (PSP-RS vs. PD) Se 60 (PSP-P vs. PD) Sp 96 (PSP-P vs. PD)
Longoni et al. (2011)	PSP-RS 10/ PSP-P 10/ PD 25/HC 24	1.5 T		MRPI	Se 100 (PSP-RS vs. PD) Sp 92 (PSP-RS vs. PD) Se 70 (PSP-P vs. PD) Sp 68 (PSP-P vs. PD)
Morelli et al. (2011a)	81 with clinically uncertain parkinsonism (of which n = 15 developed PSP and n = 11 patients developed MSA)	1.5 T	↑ MRPI in clinically uncertain parkinsonism in those patients developing PSP at follow-up compared to the other patients MRPI showed a higher accuracy in predicting PSP (92.9%) than clinical features	MRPI	Se 100 (PSP) Sp 90(PD and MSA)
Nigro et al. (2016)	PSP 88/PD 234/HC 117	1.5 T	↑ MRPI in PSP vs. PD and controls No statistical differences between automated and manual MRPI values in all groups	Automated MRPI	Se 93 (PSP) Sp 97 (vs. PD), 94 (vs. HC)
		1.5 T		Manual MRPI	Se 90 (PSP) Sp 100 (vs. PD), 94 (vs. HC)
		3.0 T		Automated MRPI	Se 97 (PSP) Sp 97 (vs. PD), 97 (vs. HC)
		3.0 T		Manual MRPI	Se 100 (PSP) Sp 100 (vs. PD) 100 (vs. HC)
Mangesius et al. (2017) (manuscript in preparation)	81 clinically uncertain parkinsonism (including n = 15 who developed PSP and n = 11 who developed MSA)	1.5 T	m_d ^c, m_a, m_a/p_a-ratio, m_d/p_d-ratio^c and MRPI showed high overall diagnostic accuracy and specificity (over 80%) in distinguishing PSP from non-PSP neurodegenerative parkinsonism	m_d	Se 89 (PSP) Sp 91 (vs. other groups), 91 (vs. PD), 92 (vs. MSA)
				m_a	Se 68 (PSP) Sp 87 (vs. other groups), 88 (vs. PD), 85 (vs. MSA)
				m_d/p_d-ratio	Se 87 (PSP) Sp 94 (vs. other groups), 94 (vs. PD), 96 (vs. MSA)
				m_a/p_a-ratio	Se 89 (PSP) Sp 85 (vs. other groups), 82 (vs. PD), 95 (vs. MSA)
				MRPI	Se 87 (PSP) Sp 86 (vs. other groups), 82 (vs. PD), 100 (vs. MSA)
Nicoletti et al. (2006a)	MSA 16 (MSA-P 13, MSA-C 3)/PD 26/HC 14	1.5 T	↓ Average MCP_d in MSA compared to PD and HC using T1-weighted sequences	MCP_d	Se 100 (MSA) 100 (vs. PD and HC)
Massey et al. (2013)	PSP 21/PD 10/MSA 10/HC 21.	3.0 T	↓ m_d ^c and reduced m_d/p_d-ratio^c in PSP compared to MSA-P, PD, and HC	m_d	Se 90 (PSP) Sp 100 (vs. PD, MSA. HC)
Massey et al. (2013)	PSP 21/PD 10/MSA 10/HC 21.	3.0 T		m_d/p_d-ratio	Se 86 (PSP) Sp 100 (vs. PD, MSA. HC)
Kim et al. (2015)	PD 82/PSP 29	3.0 T	m_d ^d and m_a/p_a ratio discriminated PSP from PD with similar discriminatory power MRPI showed lower discriminatory power	m_d	Se 86 (PSP) Sp 54 AUC = 0.76
				m_a/p_a ratio (Cosottini method) (Cosottini et al. 2007)	Se 62 (PSP) Sp 76 AUC = 0.75
				m_a/p_a ratio (Oba method) (Oba et al. 2005)	Se 72 (PSP) Sp 64 AUC = 0.75
				MRPI	Se 93 (PSP-RS) Sp 43 AUC = 0.69
Moller et al. (2017)	PD 204/PSP 106/MSA-C 21/MSA-P 60/HC 73	1.5 and 3.0 T	↓ m_a in PSP vs. all other groups ↓ p_a in MSA-C, MSA-P, and PSP vs. PD and HC. ↓ m_a/p_a in PSP vs. all other groups ↑ m_a/p_a in MSA-C and MSA-P vs. PD and PSP	m_a	AUC = 0.90 (PSP vs. PD) Se 75, Sp 82, AUC = 0.85 (PSP vs. MSA-P)
				m_a/p_a	Se 76, Sp 80, AUC = 0.84 (PSP vs. PD) Se 76, Sp 80, AUC = 0.89 (PSP vs. MSA-P)
				MRPI	Se 64, Sp 64, AUC = 0.75 (PSP vs. PD) Se 73, Sp 60, AUC = 0.80 (PSP vs. MSA-P)
Automated methods for quantitative MRI analysis
Huppertz et al. (2016)	PD 204, PSP 106, MSA-C 21, MSA-P 60	1.5 and 3.0 T	Fully automated brain volumetry combined with SVM classification allowed for automated differentiation on single-patient level Volume changes of midbrain, basal ganglia, and cerebellar peduncles had the largest relevance for classification	Atlas-based voxel-based volumetry combined with SVM classification	Classifications between the groups resulted in balanced diagnostic accuracies ≥80%
Scherfler et al. (2016)	PSP 30/MSA 40/PD 40 of whom 40 presented with a clinically uncertain parkinsonism Data were split into a training (n = 72) and a test set (n = 38)	1.5 T	Volume segmentation of subcortical brain regions followed by a machine-learning method-derived classification algorithm (i.e. C4.5 decision tree algorithm) Most discriminative regions include the volume of the midbrain, followed by cerebellar GM and putamen Diagnostic accuracy of the fully automated method for quantitative MRI analysis was 97% for the separation of PD vs. MSA or PSP, by contrast to the clinical diagnostic accuracy of 63% based on validated clinical consensus criteria at the time of MRI	Midbrain, putaminal and cerebellar GM volume	Se 90 (PSP), 100 (MSA), 100 (PD) Sp 100

APD atypical parkinsonian disorders, Se sensitivity, Sp specificity, AUC area under the curve; ↑ significant higher, ↓ significant lower, T tesla, MSA multiple system atrophy, MSA-P parkinsonian variant of MSA, MSA-C cerebellar variant of MSA, PD Parkinson’s disease, PSP progressive supranuclear palsy, HC healthy controls, MRI magnetic resonance imaging, VBM voxel-based morphometry, CC corpus callosum, WM white matter, GM grey matter, MCP middle cerebellar peduncle, SCP superior cerebellar peduncle, MRPI magnetic resonance parkinsonism index, m _d midbrain diameter, m _a midbrain area, p _d pontine basis diameter, p _a pontine area, SCP _d superior cerebellar peduncle diameter, MCP _d middle cerebellar peduncle diameter, m _d /p _d -ratio m_d to p_d ratio, m _a /p _a -ratio m_a to p_a-ratio, p _a /m _a ratio p_a to m_a ratio, MCP _d /SCP _d MCP_d to SCP_d ratio, TIV total intracranial volume, SVM support vector machine, R2* T2* relaxation rate

^aNo PD patient was classified as having MSA-P or vice versa; however, three of the PD patients were classified as having MSA-C or PSP; discrimination of patients with MSA from PSP was poor

^bDiagnostic accuracy values for planimetric measurements of regional brain structures are highest in PSP; therefore, sensitivity values were primarily given for PSP

^cThese diameters were obtained by placing elliptical regions of interest in the midbrain and pontine basis in the midsagittal slice, and obtaining the maximal measurement perpendicular to the major axis of the ellipse

^dAs obtained at the mid-mammillary-body level