PMRI Image Reconstruction Method Based on Virtual Coils and GRAPPA-enhanced Network

doi:10.11938/cjmr20253147

Abstract

Abstract:

Parallel magnetic resonance imaging (PMRI) is an imaging technique that uses multiple receiver coils for undersampling. It utilizes spatial information to supplement the insufficient gradient phase encoding and reconstructs aliasing-free images with specific algorithms to accelerate the imaging process. To address the issue of overfitting or poor generalization when using high acceleration factors with a limited number of auto calibration signals (ACS) in PMRI algorithms based on specific scans, a reconstruction method based on virtual coils and GRAPPA-enhanced networks is proposed. This method expands the sample by using virtual conjugate coils and enhances the ACS using the GRAPPA algorithm for training a nonlinear deep learning network. Experimental results show that the proposed PMRI method can effectively reduce aliasing artifacts caused by insufficient reference data, significantly improving image reconstruction quality with fewer ACS and higher acceleration factors.

Key words: parallel magnetic resonance imaging (PMRI), virtual coil concept, undersampled images, deep learning, image reconstruction

CLC Number:

TP391.4

GAO Zhaoyao, ZHANG Zhan, HU Liangliang, XU Guangyu, ZHOU Sheng, HU Yuxin, LIN Zijie, ZHOU Chao. PMRI Image Reconstruction Method Based on Virtual Coils and GRAPPA-enhanced Network[J]. Chinese Journal of Magnetic Resonance, 2025, 42(4): 390-401.

Figures/Tables 8

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Table 1

Fig. 6

Table 2

References 22

[1]	HUANG M, ZHU J L, KAO Y C, et al. Multi-coil MRI image reconstruction based on ISTAVS-Net of physical model[J]. Chinese J Magn Reson, 2024, 41(4): 418-429.
	黄敏, 朱俊琳, 考宇辰, 等. 基于物理模型的ISTAVS-Net多线圈MRI图像重建[J]. 波谱学杂志, 2024, 41(4): 418-429. doi: 10.11938/cjmr20243109
[2]	CHEN W F. Parallel magnetic resonance imaging: past, present and future[J]. Chinese Journal of Biomedical Engineering, 2005, (6): 649-654.
	陈武凡. 并行磁共振成像的回顾、现状与发展前景[J]. 中国生物医学工程学报, 2005, (6): 649-654.
[3]	LARKMAN D J, NUNES R G. Parallel magnetic resonance imaging[J]. Phys Med Biol, 2007, 52(7): R15.
[4]	PRUESSMANN K P, WEIGER M, SCHEIDEGGER M B, et al. SENSE: sensitivity encoding for fast MRI[J]. Magn Reson Med, 1999, 42(5): 952-962. pmid: 10542355
[5]	GRISWOLD M A, JAKOB P M, HEIDEMANN R M, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA)[J]. Magn Reson Med, 2002, 47(6): 1202-1210. doi: 10.1002/mrm.10171 pmid: 12111967
[6]	ZHAO T, HU X. Iterative GRAPPA (iGRAPPA) for improved parallel imaging reconstruction[J]. Magn Reson Med, 2008, 59(4): 903-907. doi: 10.1002/mrm.21370 pmid: 18383282
[7]	BLAIMER M, GUTBERLET M, KELLMAN P, et al. Virtual coil concept for improved parallel MRI employing conjugate symmetric signals[J]. Magn Reson Med, 2009, 61(1): 93-102. doi: 10.1002/mrm.21652 pmid: 19097211
[8]	WANG W T, SU S, JIA S, et al. Reconstruction of simultaneous multi-slice MRI data by combining virtual conjugate coil technology and convolutional neural network[J]. Chinese J Magn Reson, 2020, 37(4): 407-421.
	王婉婷, 苏适, 贾森, 等. 基于虚拟线圈和卷积神经网络的多层同时激发图像重建[J]. 波谱学杂志, 2020, 37(4): 407-421. doi: 10.11938/cjmr20202800
[9]	FABIAN Z, TINAZ B, SOLTANOLKOTABI M. Humus-net: Hybrid unrolled multi-scale network architecture for accelerated MRI reconstruction[J]. Adv Neural Inf Process Syst, 2022, 35: 25306-25319.
[10]	GE R, YU X, CHEN Y, et al. Tc-kanrecon: High-quality and accelerated mri reconstruction via adaptive kan mechanisms and intelligent feature scaling[J]. arxiv preprint arxiv: 2408.05705, 2024.
[11]	AKÇAKAYA M, MOELLER S, WEINGÄRTNER S, et al. Scan-specific robust artificial-neural-networks for k-space interpolation (RAKI) reconstruction: database-free deep learning for fast imaging[J]. Magn Reson Med, 2019, 81(1): 439-453. doi: 10.1002/mrm.27420 pmid: 30277269
[12]	ZHANG C, MOELLER S, DEMIREL O B, et al. Residual RAKI: A hybrid linear and non-linear approach for scan-specific k-space deep learning[J]. NeuroImage, 2022, 256: 119248. doi: 10.1016/j.neuroimage.2022.119248
[13]	ZHENG H R, WU Y, HE Q, et al. Fast and high-resolution magnetic resonance imaging on high field system[J]. Life Science Instruments, 2018, 16(Z1): 29-44+54.
	郑海荣, 吴垠, 贺强, 等. 基于高场磁共振的快速高分辨成像[J]. 生命科学仪器, 2018, 16(Z1): 29-44+54.
[14]	KEBAILI A, LAPUYADE-LAHORGUE J, RUAN S. Deep learning approaches for data augmentation in medical imaging: a review[J]. J Imaging, 2023, 9(4): 81. doi: 10.3390/jimaging9040081
[15]	YU X, WANG J, HONG Q Q, et al. Transfer learning for medical images analyses: A survey[J]. Neurocomputing, 2022, 489: 230-254. doi: 10.1016/j.neucom.2021.08.159
[16]	CHEN Y, YANG X H, WEI Z, et al. Generative adversarial networks in medical image augmentation: a review[J]. Comput Biol Med, 2022, 144: 105382. doi: 10.1016/j.compbiomed.2022.105382
[17]	GOODFELLOW I, POUGET-ABADIE J. Generative adversarial networks (GANs) for low-data MRI reconstruction[J]. Nat Mach Intell, 2019, 1(5): 236-245. doi: 10.1038/s42256-019-0052-1
[18]	DENCK J, GUEHRING J, MAIER A, et al. Enhanced magnetic resonance image synthesis with contrast-aware generative adversarial networks[J]. J Imaging, 2021, 7(8): 133. doi: 10.3390/jimaging7080133
[19]	NING X Z, HUANG Z, CHEN X Q, et al. Research on transformer super-resolution reconstruction algorithmfor ultrafast spatiotemporal encoding magnetic resonance imaging[J]. Chinese J Magn Reson, 2024, 41(4): 454-468.
	宁欣宙, 黄臻, 陈西曲, 等. 用于超快时空编码MRI的Transformer超分辨率重建算法研究[J]. 波谱学杂志, 2024, 41(4): 454-468. doi: 10.11938/cjmr20243110
[20]	ZHANG C, HOSSEINI S A H, WEINGÄRTNER S, et al. Optimized fast GPU implementation of robust artificial-neural-networks for k-space interpolation (RAKI) reconstruction[J]. PloS one, 2019, 14(10): e0223315.
[21]	ROEMER P B, EDELSTEIN W A, HAYES C E, et al. The NMR phased array[J]. Magn Reson Med, 1990, 16(2): 192-225. doi: 10.1002/mrm.1910160203 pmid: 2266841
[22]	WALSH D O, GMITRO A F, MARCELLIN M W. Adaptive reconstruction of phased array MR imagery[J]. Magn Reson Med, 2015, 43(5): 682-690. doi: 10.1002/(ISSN)1522-2594

ACS	R	算法	NRMSE	SSIM	PSNR
16	4	ZF	0.3311	0.8475	23.8473
		GRAPPA	0.0240	0.9248	35.2355
		RAKI	0.0757	0.9282	30.2510
		rRAKI	0.0702	0.9391	30.5825
		本文方法	0.0108	0.9679	38.6888
10	5	ZF	0.6210	0.7924	21.1159
		GRAPPA	0.0787	0.8544	30.0844
		RAKI	0.2532	0.7941	25.0115
		rRAKI	0.1954	0.7880	26.1367
		本文方法	0.0226	0.9547	35.0208
18	6	ZF	0.2651	0.8599	24.2725
		GRAPPA	0.1346	0.8069	27.7540
		RAKI	0.0714	0.9071	29.9671
		rRAKI	0.0967	0.8927	28.6522
		本文方法	0.0229	0.9518	34.8806
24	8	ZF	0.3113	0.8669	24.1142
		GRAPPA	0.1516	0.8021	27.2378
		RAKI	0.1941	0.8215	26.1653
		rRAKI	0.0984	0.8536	29.1153
		本文方法	0.0599	0.9205	31.2673
30	10	ZF	0.3365	0.8723	23.7771
		GRAPPA	0.1391	0.8205	27.6127
		RAKI	0.2295	0.8084	25.4376
		rRAKI	0.1906	0.7841	26.2441
		本文方法	0.0982	0.9323	29.1251
30	15	ZF	0.3021	0.8749	24.1621
		GRAPPA	0.1606	0.8219	26.9897
		RAKI	0.2651	0.8748	24.7297
		rRAKI	0.1834	0.8605	26.3309
		本文方法	0.0781	0.9231	30.0419

ACS	R	残差结构	虚拟共轭线圈生成	迭代训练方法	NRMSE	SSIM	PSNR
10	2	×	×	×	0.0032	0.9847	42.8324
		√	×	×	0.0031	0.9850	42.8832
		√	√	×	0.0024	0.9855	43.9197
		√	√	√	0.0020	0.9797	44.7016
10	5	×	×	×	0.1985	0.7845	24.9643
		√	×	×	0.1857	0.7850	25.0925
		√	√	×	0.1684	0.7834	25.5169
		√	√	√	0.0136	0.9512	36.4516
16	4	×	×	×	0.1456	0.9185	26.1477
		√	×	×	0.0962	0.9460	27.9456
		√	√	×	0.0202	0.9693	34.7181
		√	√	√	0.0084	0.9489	38.5150
24	8	×	×	×	0.2910	0.7816	23.1409
		√	×	×	0.2798	0.7844	23.3122
		√	√	×	0.2347	0.7927	24.0741
		√	√	√	0.0349	0.9246	32.3490
30	10	×	×	×	0.4402	0.6674	21.3433
		√	×	×	0.2402	0.7993	23.9745
		√	√	×	0.2131	0.7791	24.4926
		√	√	√	0.1285	0.9233	26.6901