Automatic segmentation of prostate zonal anatomy on MRI: a systematic review of the literature

Wu, Carine; Montagne, Sarah; Hamzaoui, Dimitri; Ayache, Nicholas; Delingette, Hervé; Renard-Penna, Raphaële

doi:10.1186/s13244-022-01340-2

Insights into Imaging

Table 2 Overview of types of databases used with training, validation and test sets distribution

From: Automatic segmentation of prostate zonal anatomy on MRI: a systematic review of the literature

First author, year of publication	Inclusion criteria	Presence of PCa	Number of patients (total)	Training			Validation		Test
First author, year of publication	Inclusion criteria	Presence of PCa	Number of patients (total)	Total	Public data	In-house data	Cross-validation	Validation data	Total	Public data	In-house data
Cuocolo et al. [43]	✓	✓	204	79	79^(A)	0	Fivefold	20	105	105^(A)	0
Bardis et al. [42]	✓	✓	242	146	0	145	0	48	48	0	48
Lai et al. [41]	✓	✓	115	80	80^(A)	0	Fivefold	20	15	15^(A)	0
Nai et al. [18]	✓	✓	160	120	120^(A)	0	0	20	20	20^(A)	0
Sanford et al. [40]	✓	✓	1054	518 + 162^§	0	680	0	130 + 42^§	202	0	202
Aldoj et al. [39]	✓	✓	188	106	106^(A)	0	Fourfold	35	47	20^(A)	0
Zavala-Romero et al. [6]	✓	✓	550	198 or 297	297^(A)	198	0	0	variable	33^(A)	22
Lee et al. [38]	✓	✓	330	260 (for WG) or 162 (for TZ)	0	260 (for WG) or 162 (for TZ)	0	0	70 (for WG) or 50 (for TZ)	0	50
Liu et al. [37]	✓	✓	351	218	218^(A)	0	0	45	92	45^(A)	47
Qin et al. [36]	×	✓	240	162 + 45	45 ^(B)	162	0	0	33	15 ^(B)	18
Motamed et al. [35]	×	Unknown	681	291 (source) + variable (target)	0	406	0	97	145 (source) + 33 (target)	0	178
Zabihollahy et al. [13]	✓	✓	225	80	0	80	0	20	125	0	125
Padgett et al. [8]	✓	✓	61	Variable	0	Variable	0	0	Variable	0	1
Rundo et al. [15]¹	×	✓	80	Variable	Variable^(B)	Variable	0	0	Variable	Variable^(B)	Variable
Meyer et al. [34]	✓	✓	98	58	58^(A)	0	Fourfold	20	20	20^(A)	0
Liu et al. [16]	✓	✓	359	200	200^(A)	0	Fivefold	50	110	63^(A)	46
Rundo et al. [33]²	×	✓	40 ^†	Variable ^†	^†(C)	Variable	0	0	Variable	0	Variable
Hambarde et al. [32]	×	Unknown	52	42	0	42	0	0	10	0	10
Jensen et al. [31]	✓	✓	40	32	0	32	Fivefold	2	8	0	8
Khan et al. [17]	×	✓	80	35	35^(B)	0	0	15	30	30^(B)	0
Cheng et al. [30]	×*	✓	225	116 +/−^†	8^(A)	108	0	0	Variable	Variable^(A+C)	27
Zhu et al. [12]	✓	✓	163	76	0	76	0	36	51	0	51
Mooij et al. [29]	0	Unknown	53	36	0	36	Fivefold	9	8	0	8
Can et al. [28]	0	Unknown	29	12	12^(B)	0	0	7	10	10^(B)	0
Clark et al. [14]	×*	✓	154	115	78^(C)	37	0	0	38	12^(C)	26
Chilali et al. [9]	✓	✓	55	30	30 (Prostatlas)	0	0	0	25	13^(C)	12
Makni et al. [10]	✓	✓	31	? (simulated images)	0	0	0	0	31	0	31
Chi et al. [27]	✓	Unknown	8	4	0	4	0	0	4	0	4
Toth et al. [26]	✓	✓	40	Variable	0	Variable	0	0	Variable	0	Variable
Litjens et al. [7]	×	Unknown	48	48	0	47	0	0	1	0	1
Moschidis and Graham [25]	✓	×	22	Variable	0	Variable	0	0	Variable	0	Variable
Yin et al. [24]	×	✓	522 (images)	261 (images)	0	261 (images)	Fivefold	52 (images)	261 (images)	0	261 (images)
Makni et al. [23]	✓	✓	31	?	0	0	0	0	31	0	31

*Not specified for in-house data
^†+/− 50 patients from PROMISE12 dataset used for pre-training of WG segmentation
^§Pre-training data + data for transfer learning
^(A)Public data used is PROSTATE-X
^(B)Public data used is NCI-ISBI
^(C)Public data used is PROMISE12
¹Rundo et al., USE-Net: incorporating Squeeze-and-Excitation blocks into U-Net for prostate zonal segmentation
of multi-institutional MRI datasets [30]
²Rundo et al., CNN-based Prostate Zonal Segmentation on T2-weighted MR Images: A Cross-dataset Study[27]
PCa prostate cancer, WG whole gland, TZ transition zone

Back to article page