overview_sct

Overview sCT

Repository to collect all the references on generation of synthetic computed tomography (sCT) with deep learning/convolutional networks. Generated from Spadea MF & Maspero M et al. Med. Phys. 2021 (in press), https://doi.org/10.1002/mp.15150, preprint at: http://arxiv.org/abs/2102.02734. This page is available at: https://matteomaspero.github.io/overview_sct/ and its Github repository is: https://github.com/matteomaspero/overview_sct/.

By the end of 2021 the table will be updated, so far all the paper up to Jan 2021 should be included, if not; please, feel free to contribute!

License

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

In short, this means that anyone, even a commercial entity may re-use the content of this page as long as it will cite our paper and the source.

Contributors

Maspero M is the owner administrator of the project. Spadea Maria Francesca and Paolo Zaffino greatly contributed to the data collection for the publication.

The following added content: …

Make the table sortable

If you wish to extend the interaction with the tables, you can sort the columns in your favourite browser, as described here. We have tested on Chrome the following combination:

install Tampermonkey;
install Github Sort Content;

If you see any typos/mistakes

Ok, we did our best, but you got us! That can happen, but you can contribute to the accuracy of the information in this page by proposing a commit, after having forked this repository. Matteo will keep monitored the repository and update it every two months. If you wish to see some changes, just reach out; we are more than happy to hear your comments/suggestions.

MRI without dose evaluation

Tumour site	train	val	test	x-fold	field [T]	sequence	conf	arch	Reg	MAE [HU]	PSNR [dB]	SSIM	others	reference	pub date
Abdomen	10^v		10	LoO		mDixon	2D pair	GAN^*	def	61±3			CC	Xu2019	2019-11-06
Abdomen	160			LoO	n.a.	n.a.	2D pair	GAN^*	rig	5.1±0.5		.90±.43	(F/M)SIM IS …	Xu2020	2020-05-08
Brain	18			6x	1.5	3D T1 GRE	2D pair	U-net	rig	85±17			MSE ME	Han2017	2017-02-13
Brain	16			LoO	n.a.	T1	2.5Dp patch	CNN+	rig	85±9	27.3±1.1			Xiang2018	2018-03-30
Brain	15			x5	1.0	T1 Gd	2D pair	GAN^*	def	89±10	26.6±1.2	.83±.03	tissues	Emami2018	2018-06-14
Brain	98CT 84MR		10		3	3D T2	2D pair/unp	GAN	aff	19±3	65.4±0.9	.25±.01		Jin2019	2019-05-22
Brain	24			LoO	n.a.	T1	3Dp pair	GAN	rig	56±9	26.6±2.3		NCC, HD body	Lei2019	2019-05-21
Brain	33 160			LoO	n.a. n.a.	T1^b n.a.	2D	GAN	yes no	9.0±0.8 5.1±0.5		.75±0.77 .90±.43	FSIM MSIM IS SWD FID experts	Xu2020	2020-05-08
Brain	28^t	2	15		1.5	n.a.	2D	GAN^*	aff	134±12	24.0±0.9	.76±.02		Yang2020	2020-11-30
Brain	28		6		1.5	T2	2D pair 2D unp	U-net GAN	rig	65±4 94±6	28.8±0.6 26.3±0.6	.972±.004 .955±0.007	same metrics for synth MRI	Li2020	2020-11-05
Brain	81		11	8x	1.5	3D T1 GRE 3D T1 GRE Gd 2D T2 SE 2D T2 FLAIR	2D	U-net	aff	45.4±8.5 44.6±7.4 45.7±8.8 51.2±4.5	43.0±2.0 43.4±1.2 43.4±1.2 44.9±1.2	.65±.05 .63±.03 .64±.03 .61±.04	metrics for air bone, soft tissue, DSC bones	Massa2020	2020-11-27
H&N	23		10		1.5	T2	2D pair	U-net	def	131±24			MAE, ME soft tis./bone	Wang2019	2019-11-29
H&N	28	4		8x	1.5	2D T1±Gd, T2	2D pair	GAN	aff	75.7±14.6	29.1±1.6	.92±.02	DSC, MAE on bone	Tie2020	2020-02-03
H&N	60	30		3	T1	2D unp	GAN	n.a.	19.6±0.7	62.4±0.5	.78±0.2			Kearney2020	2020-03-25
H&N	7		8	LoO	1.5	3D T1, T2	2D pair	GAN	def	83±49			ME	Largent2020	2020-03-14
H&N	10			LoO	1.5	3D T1, T2	2D pair	GAN^*	def	42-62			RMSE, CC	Qian2020	2020-03-10
H&N	32		8	5x	3	3D UTE	2D pair	U-net	def	104±21			DSC, spatial corr	Su2020	2020-10-06
Prostate	16 22			LoO	n.a.	T1	2.5Dp pair	CNN+	rig	85±9 43±2	27.3±1.1 33.5±0.8			Xiang2018	2018-03-30
Pelvis	20			LoO	n.a.	3D T2	3Dp pair	GAN^*	rig	51±16	24.5±2.6		NCC, Hausdorff on body	Lei2019	2019-05-21
Prostate	20			5x	1.5	2D T1 TSE	2D pair 3D p pair	U-net	def	41±5 38±5			DSC bone	Fu2019	2019-06-20
Pelvis human Pelvis canine	27 18			3x	3 1.5	3D T1 GRE mDixon	3Dp pair	U-net	def	32±8	36.5±1.6		MAE/DSC bone surf dist<0.5 mm	Florkow2019	2019-10-08
Pelvis	15		4	5x	3	3D T2	2D pair	CNN U-net	def	38±6 43±9	29.5±1.2 28.2±1.6	.96±.01 .95±.01	ME, PCC	Bahrami2020	2020-07-30
Pelvis	100				3	2D T2 FSE	2D unp	GAN	No				FID	Fetty2020	2020-11
Breast	14		2	LoO	n.a.	n.a.	2D	U-net¹	def				DSC .74-.76	Jeon2019	2019-12-31

Super/subscripts
^vvolunteers, not patients; ¹to segment CT into 5-classes; ^amultiple combinations of Dixon images was investigated but omitted here; ^bdataset from http://www.med.harvard.edu/AANLIB/ ; ^trobustenss to training size was investigated;^*comparison with other architecture has been provided; ⁺trained in 2D on multiple view and aggregated after inference;
Abbreviations
H&N=head and neck ; val=validation; x-fold=cross-fold ;conf=configuration; arch=architecture; GRE=gradient echo; (T)SE=(turbo) spin-echo, mDixon = multi-contrast Dixon reconstruction; LoO=leave-one-out; (R)MSE=(root) meas squared error; ME=mean error; DSC=dice score coefficient; (N)CC=normalized cross correlation; FSIM, MSIM, IS, SWD, FID, PCC look up the references ;)

MRI-to-sCT with dose evaluation

Tumour site	train	val	test	x-fold	field [T]	sequence	conf	arch	pair reg	MAE [HU]	PSNR [dB]	others	Plan	DD [%]	GPR [%]	DVH	others	reference	pub date
Liver	21			LoO	3	3D T1 GRE	3D pair	GAN	def	73±18	22.7±3.6	NCC	p		99.4±1.0³	<1%	range γ₂ γ₁	LiuY2019	2019-06-16
Abdomen	12			4x	0.3 1.5	GRE	2D pair br> 2D unp	GAN^*	def	90±19² 94±30²	27.4±1.6 27.2±2.2		x+B₀	<±0.6 <±0.6	98.7±1.5% 98.5±1.6%	<±0.15	γ₃	Fu2020	2020-01-31
Abdomen	46		31	3x	3	3D T1 GRE	2.5D pair	U-net	syn rig	79±18		MAE, ME organs	x			<2Gy		Liu2020	2020-06-11
Abdomen kids	54	18	12	3x	1.5 3	3D T1 GRE, T2 TSE	3Dp pair	U-net	def	62±13	30.0±1.8	ME, DSC tissues	x p	<0.1 <0.5	99.7±0.3² 96.2±4.0²	<2% <3%	beam depth	Florkow2020	2020-10-07
Abdomen	39		19		0.35	GRE	2D pair	U-net	def	79±18		ME tissues	x+B₀	<0.1	98.7±1.1²	<2.5%	γ₃ γ₁	Cusumano2020	2020-10-17
Brain	26			2x	1.5	3D T1 GRE	m2D+ pair	CNN	rig	67±11		ME, tissues DSC, dist body	x	-0.1±0.3	99.8±0.7²		beam γ₃ depth γ₁	Dinkla2018	2018-11
Brain	40		10		1.5	3D T1 GRE Gd	2D pair	CNN	def	75±23		DSC	x	<0.2±0.5	99.2³			LiuF2019	2019-03-12
Brain	54	9	14	5x	1.5	2D T1 SE Gd	2D pair	GAN	rig	47±11		each fold	x	-0.7±0.5	99.2±0.8²	<1%	2D/3D γ₃ γ₁	Kazemifar2019	2019-04-11
Brain	55	28	4		1.5	3D T1 GRE	2D pair 3Dp pair	U-net	rig	116±26 137±32		ME	x p		>98^2,98±2² >98^2,97±3²		range γ₁	Neppl2019	2019-07-04
Brain	25	2	25		1.5	3D T1 GRE	3Dp pair	GAN	rig	55±7		ME DSC	x	<2	98.4±3.5²	<1.65%	range γ₃ γ₁	Shafai2019	2019-09-30
Brain	47		13	5x	3	T1	2D pair	U-net	rig	81±15		ME air, tissues	x	2.3±0.1			align CBCT<0.5mm	Gupta2019	2019-10-25
Brain	12	2	1	LoO	3	3D T1 GRE	2D+ pair	U-net	rig	54±7		ME, DSC	p	0.00±0.01			range tissues	Spadea2019	2019-11-01
Brain	15			5x		T1, T2 FLAIR^c	2Dp pair	GAN	def	108±24		tissues	x	0.7	99.2±1.0²	<1%	beam depth γ₃ γ₁	Koike2019	2019-12-10
Brain	66		11	5x	1.5	2D T1 SE Gd	2D unp	GAN	rig	78±11			p	0.3±0.3	99.2±1.0²	<3%	beam γ₃ depth γ₁	Kazemifar2020	2020-03-26
Brain kids	30^t	10	20	3x	1.5 3	3D T1 GRE±Gd	2D+^* pair	GAN^*	rig	61±14	26.7±1.9	ME DSC SSIM	x p	-0.1±0.3 0.1±0.4	99.5±0.8² 99.6±1.1²	<1% <3%	beam depth γ₃	Maspero2020	2020-10-23
Brain	242^m,t	81	79		3 1.5	3D T1 GRE±Gd	3Dp pair	CNN U-net	def	81±22 90±21		tissues	x	0.13±0.13 0.31±0.18	99.6±0.3² 99.4±0.5²	<±0.15	γ₃	Andres2020	2020-11
Brain	26	15	12		1.0	T1 Gd	2D	GAN	def			bone	x	<±1		<1.5%		Liu2021	2021-01-07
Prostate Rectum Cervix	32		27 18 14		3 1.5 1.5/3	3D T1 GRE mDixon	2D pair	GAN	rig	60±6 56±5 59±6		ME	x	-0.3±0.4 -0.3±0.5 -0.1±0.3^a	99.4±0.6³ 98.5±1.1³ 99.6±1.9³	<1%	γ2</sub></sub>	Maspero2018	2018-09-10
Prostate	36		15		3	T2 TSE	2D pair	U-net	def	30±5		ME tissues	x	0.16±0.09	99.4²	<0.2Gy	γ₃ γ₁	Chen2018	2018-10-20
Prostate	39			4x	3	3D T2	2D pair	U-net	def	33±8		ME, DSC dist body	x	-0.01±0.64	98.5±0.7²	<3%	γ₃ γ₁	Arabi2018	2018-10-14
Prostate	17			LoO	1.5	T2	3D patch	GAN^*	No	51±17	24.2±2.5	NCC, bone:dist, uniform	p	-0.07±0.07	98±6²	<1%	range peak γ₃ γ₁	LiuY2019b	2019-10-21
Prostate	25		14	3x	3	3D T2 TSE	2D	U-net^* GAN^*	def	34±8 34±8		ME tissues	x	<1% <1%	99.2±1¹ 99.1±1¹	<1%		Largent2019	2019-12-1
Pelvis	11m		8		3 1.5	T2 TSE	2D	GAN^*	def	49±6		ME organs	x	0.7±0.4	99.2±1.0²	<1.5%		Boni2020	2020-04-02
Pelvis	26	15	10+19^m		0.35 1.5/3	3D T2	2.5D	GAN^*	def	41±4	31.4±1	ME MSE bone	x	<±1		<1.5%		Fetty2020	2020-05-22
Pelvis	39		14		0.35	GRE	2D pair	U-net	def	54±12		ME tissues	x+B₀	<0.5	99.0±0.7²	<1%	γ₃ γ₁	Cusumano2020	2020-10-17
Rectum	46m		44		1.5	3D T2	2D	GAN	def	35±7		ME bone	x	<±0.8	99.8±0.1²	<1%	γ₃ γ₁	Bird2020	2020-11-29
H&N	34			3x	1.5	3D T2 TSE	3Dp pair	U-net	def	75±9		ME DSC bone	x	-0.07±0.22	95.6±2.9²		γ₃	Dinkla2019	2019-06-17
H&N	15		12		3	T1 GRE	2Dp pair^*	GAN^*	def		68±2	SSIM RMSE	p	<0.5	<98²	<0.5		Klages2019	2019-11-16
H&N	30		15		3	T1±Gd T2 TSE^c	2D pair	GAN^* U-net	rig	70±12 71±12	29.4±1.3 29.2±1.3	SSIM DSC, DRR	p	-0.3±0.2 -0.2±0.2	97.8±0.9² 97.6±1.3²			Qi2020	2020-02-06
H&N	135^t	10	28		3	3D T1 GRE	2D pair unp	GAN^*	def	70±9 101±8		ME, DSC tissues	x	-0.1±0.3 0.1±0.4	98.7±1.0² 98.5±1.1²	<1.5% <1.5%	beam depth	Peng2020	2020-07-03
H&N	27			3x	3	3D T1 GRE	2D+ pair	GAN	def	65±4		ME	p	<±0.2	93.5±3.4	<1.5%	NTCP DSC RS γ₃	Thummerer2020	2020-11-27
Breast	12^t		18	LtO	1.5	3D GRE mDixon	2D + patch	GAN^*	def	94±11 103±15		NCC	p	<0.5	98.4±3.5²		DRR dist bone	Olberg2019	2019-11-16

Super/subscripts
^*comparison with other architecture has been provided; ³γ_3%,3mm = γ₃; ²γ_2%,2mm = γ₂; ¹ γ_1%,1mm = γ₁; ^trobustenss to training size was investigated; ⁺trained in 2D on multiple view and aggregated after inference; ^c multiple combinations (also ± Dixon reconstruction, where present) of the sequences were investigated but omitted; ^m data from multiple centers
Abbreviations
H&N=head and neck ; val=validation; x-fold=cross-fold ;conf=configuration; arch=architecture; GRE=gradient echo; (T)SE=(turbo) spin-echo, mDixon = multi-contrast Dixon reconstruction; LoO=leave-one-out; (R)MSE=(root) meas squared error; ME=mean error; DSC=dice score coefficient; (N)CC=normalized cross correlation;

CBCT

Tumour site	train	val	test	x-fold	conf	arch	pair reg	MAE [HU]	PSNR [dB]	SSIM	others	Plan	DD [%]	DPR [%]	GPR [%]	DVH	others	reference	pub date
Pancreas	30			LoO	3Dp pair	GAN^*	def	56.89±13.84	28.80±2.46	.71±.032	NCC	SNU	x			<1Gy		Liu2020	2020-03-06
Brain Pelvis	24 20			LoO	3D patch	GAN	rig	13±2 16±5	37.5±2.3 30.7±3.7		NCC SNU	No						Harms2019	2019-06-17
Prostate	16		4	5x	2D pair	U-net	def		50.9	.967	SNU RMSE	No						Kida2018	2018-04-29
Prostate	27	7	8		2D pair	U-net^*	def	88			ME	x p		>98.4¹ 88.5³	99.5² >96.5²		γ1 DPR₂	Landry2019	2019-01-24
Prostate	18		8	4x	2D ens unp	GAN	No	rig	87±5			ME	x p	99.9±0.3² 80.5±5²	95.9±2.0²	<±1.5% <1%	DPR¹ DPR³ RS γ₃	Kurz2019	2019-11-15
Prostate	16		4		2D pair	GAN^*	rig				SelfSSIM tissues	No						Kida2019	2019-12-16
Pelvis H&N	135	15	15 10	10x	2.5D pair	GAN^*	def	24±5 24±4	20.1±3.4 22.8±3.4			x, p	<1%				RS	Zhang2020	2020-12-01
H&N Lung Breast	15 15 15	8 8 8	10 10 10		2D	GAN^*	No rig	53±12 83±10 66±18	30.5±2.2 28.5±1.6 29.0±2.1	.81±.04 .78±.04 .76±.02	ME	x	0.1±0.5 0.2±0.9 0.1±0.4		97.8±1² 94.9±3² 92±8²	<2%	γ₃	Maspero2020	2020-04-29
H&N	81	9	20		2D	GAN^*	No def	29.85±4.94	30.65±1.36	.85±.03	RMSE phantom	x			98.4±1.7² 96.3±3.6¹			Liang2019	2019-06-10
Nasophar	50	10	10		2D	U-net	rig	6-27			ME organs	x	0.2±0.1		95.5±1.6¹	<1%		Li2019	2019-07-16
H&N	30	7	7		2D	U-net^*	rig	18.98	33.26	0.8911	RMSE tissues	No						Chen2019	2019-12-18
H&N	30		14		2D	GAN	def	82±11			ME tissues	x	91.0±5.3²		<1Gy	<1%		Barateau2020	2020-07-12
H&N	22		11	3x	2D+	U-net	def	36±6			ME DSC SNU	p	-0.1±0.3		98.1±1.2²		RS γ₃	Thummerer2020	2020-04-27
H&N	50^t		10		2.5D	U-net	rig	49		.85	SNR	No						Yuan2020	2020-01-24
H&N	23			LoO	3D patch	GAN^*	rig		24.3±1.4	.80±.05	stop power	p			88.4²	<1%	γ₃ γ₂	Harms2019	2019-06-17
H&N Thorax Pelvis	25 53 205			15 15 15	2D	GAN	def	77±13 94±32 42±5			ME DSC HD tissues	x		91.5±4.3² 76.7±17.3² 88.9±9.3²	95.0±2.4² 93.8±5.9² 98.5±1.7²	<2.4 <2.6 <1	γ₃	Eckl2020	2020-11-24

PET

Region	train	val	test	x-fold	field [T]	image contrast	conf	arch	pair reg	MAE [HU]	DSC	tracer	PET_err [%]	others	reference	pub date
Pelvis	10		16		3	Dixon ZTE	3D patch	U-net	def			¹8F-FDG 68Ga-PSMA		RMSE SUV diff	Leynes2017	2017-05-01
Head	30		10		1.5	T1 GRE Gd	2D	auto-enc	def	.971±.005^a .936±.011^s	.803±.021^b	n.a.	-0.7±1.1		Liu2018	2017-10-19
Pelvis	12		6		3	T1 GRE T2 TSE	3D	CNN¹	def		.98±.01^s .79±.03^b .49±.17^a	¹8F-FDG		RMSE	Bradshaw2018	2018-09
Head	30p+6		8			UTE mDixon	2D	U-net¹	def		.96±.006^b	¹8F-FDG	<1%		Jang2018	2018-5-15
H&N	32 12		8 2	5 7	3	Dixon±ZTE	2D	U-net	rig	13.8±1.4 12.6±1.5	.76±.04^b .80±.04^b	¹8F-FDG	<3		Gong2018	2018-06-13
Pelvis	15		4	4	3	T1 GRE Dixon	2D	U-net	def			¹8F-FDG 1.8±2.4	1.7±2.0f 1.8±2.4^s 3.8±3.9^b	mu-map diff	Torrado2019	2018-08-30
Head	23		47		3	ZTE	3D patch	U-net	def		.81±.03^b	¹8F-FDG	-0.2±5.6	Jac	Blanc-Durand2019	2019-10-07
Head kids	60		19	4	3	T1 GRE mDixon, UTE	3D	U-net	rig		.90±.07j	¹8F-FET		biol tumor vol, SUV	Ladefoged2019	2019-01-07
Head	44	11	11		1.5	T1 GRE	2.5D	U-net	rig			¹1C-WAY ¹1C-DASB	-0.49±1.7	synt mu-map, kin anal	Spuhler2019	2019-08-30
Head	40			2	3	3D T1 GRE	3D patch	GAN	def	101±40 302±79^b 407±228^a 8±4^s	.80±.07^b	¹8F-FDG	3.2±3.4 1.2±13.8^b 3.2±13.6^s 3.2±13.6^a	rel vol dif surf dist ME RMSE PSNR SSIM SUV	Arabi2019	2019-07-01
Prostate	18		10		3	Dixon	2D	GAN^*	def			68Ga-PSMA	2.4±0.5	SSIM SUV	Pozaruk2020	2020-05-11
Head	35			5	3	3D T1 GRE mDixon+UTEc	2.5D	U-net	rig	11.94±0.01	.87±.03^b	¹1C-PiB ¹8F-MK-6240	<2%		Gong2020	2020-10-27
Head	32			4	3	Dixon	3D patch	GAN^*	def	16±2%	.74±.05^b	¹8F-FDG	-1.0±13	SUV	Gong2020	2020-07-03
Thorax	14			LoO	3	Dixon	2D	GAN^*	No def	68±10		¹8F-NaF		PSNR SSIM RMSE	Baydoun2020	2020-12

Body	100		28	PET, no att corrected	2D	U-net	Yⁱ	111±16	.94±.01^b	¹8F-FDG	-0.6±2.0%	abs err	Liu2018	2018-11-12
Body	100		25	PET, no att corrected	2.5D	GAN	Yⁱ			¹8F-FDG	-0.8±8.6%	SUV ME	Armanious2020	2020-05-24
Body	80		39	PET, no att corrected	3D	GAN	Yⁱ	109±19	.87±.03^b	¹8F-FDG	0.1<3.0%	NCC PSNR ME	Dong2019	2019-11-4

Super/subscripts
^*comparison with other architecture has been provided; ^ain air or bowel gas; ^bin the bony structures; ^sin the soft tissue; ^f in the fatty tissue; ^w in water; ^jexpressed in terms of Jaccard index and not DSC; ⁱintrinsically registered: PET-CT data; ^pdata from another MRI sequence used as pre-training; ^c multiple combinations (also ± Dixon reconstruction, where present) of the sequences were investigated but omitted; Abbreviations
H&N=head and neck ; val=validation; x-fold=cross-fold ;conf=configuration; arch=architecture; GRE=gradient echo; (T)SE=(turbo) spin-echo, mDixon = multi-contrast Dixon reconstruction; LoO=leave-one-out; (R)MSE=(root) meas squared error; ME=mean error; DSC=dice score coefficient; (N)CC=normalized cross correlation; paed=paediatric