Study on deep learning reconstruction technology in improving image quality of pituitary neuroendocrine tumors in coronal T1WI magnetic resonance image

doi:10.16150/j.1671-2870.2024.06.006

Abstract

Abstract:

Objective To investigate the role of thin-layer pituitary T1WI fat suppression (FS) coronal enhanced sequences based on deep learning (DL) reconstruction technology in improving image quality for pituitary neuroendocrine tumors. Methods From June 2023 to June 2024, 46 patients diagnosed or suspected of having pituitary lesions were prospectively and consecutively enrolled, with a total of 40 pituitary neuroendocrine tumor lesions identified. All patients underwent thin-layer pituitary DL T1WI FS coronal enhanced scanning. Original reconstruction (OR) images without DL application were retained, and the images were divided into DL and OR groups according to the reconstruction method. Two neuroradiologists, using a double-blind method, subjectively evaluated (using the five-point Likert scale) the image quality of the two groups in six aspects: uniformity, sharpness, artifact, pituitary structure identification, lesion identification, and overall quality. Objective evaluation included measuring and calculating the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of pituitary neuroendocrine tumor and tumor-free area. The differences in image quality scores between the two groups were compared using the Wilcoxon rank-sum test. The intra-class correlation coefficient (ICC) was used to assess the consistency of subjective and objective image measurement results of both doctors. Results The ICC values of subjective and objective image quality scores between the DL and OR groups were all greater than 0.81, indicating extremely high consistency. In terms of subjective image quality evaluation, the image uniformity of the DL and OR groups was 4.33 (3, 5) and 3.73 (3, 4), respectively. Sharpness was 4.25 (3, 5) and 3.50 (3, 4), artifact was 4.35 (4, 5) and 2.95 (2, 4), pituitary structure identification was 4.38 (3, 5) and 3.35 (2, 5), lesion identification was 4.6 (3, 5) and 3.15 (2, 4), and overall quality was 4.30 (4, 5) and 2.63 (2, 3), respectively. The DL group showed higher scores than the OR group, and the differences were statistically significant (all P<0.001). For objective image quality evaluation, the SNRs of pituitary tumors in the DL and OR groups were 26.96 (18.10, 34.15) and 16.51 (11.24, 20.65), respectively, while the CNRs were 11.30 (6.74, 19.72) and 4.34 (3.07, 6.00), respectively. The SNRs of the pituitary in the DL and OR groups were 38.36 (31.93, 47.03) and 17.02 (15.49, 20.51), while the CNRs were 29.89 (23.28, 39.75) and 18.44 (16.61, 24.56), respectively. The DL group had higher scores than the OR group, and the differences were statistically significant (all P<0.001). The subjective and objective indicators evaluated by the two doctors showed good consistency. Conclusions The T1WI FS coronal enhanced sequences based on DL can significantly improve the image quality of pituitary neuroendocrine tumors and enhance the SNR and CNR of the images while ensuring the spatial resolution of the images, providing accurate imaging support for clinical diagnosis and treatment.

Key words: Pituitary, Neuroendocrine tumor, Deep learning, Image reconstruction, Magnetic resonance imaging

CLC Number:

R736.4

ZHANG Huihui, FANG Shu, WU Mengxiong, LIU Fangtao, HE Naying, DONG Haipeng, YAN Fuhua. Study on deep learning reconstruction technology in improving image quality of pituitary neuroendocrine tumors in coronal T1WI magnetic resonance image[J]. Journal of Diagnostics Concepts & Practice, 2024, 23(06): 594-601.

Figures/Tables 7

Table 1

Table 2

Table 3

Figure 1

Figure 2

Figure 3

Table 4

References 34

[1]	DI IEVA A, ROTONDO F, SYRO L V, et al. Aggressive pituitary adenomas--diagnosis and emerging treatments[J]. Nat Rev Endocrinol, 2014, 10(7):423-435. doi: 10.1038/nrendo.2014.64 pmid: 24821329
[2]	WANG X, DAI Y, LIN H, et al. Shape and texture analyses based on conventional MRI for the preoperative prediction of the aggressiveness of pituitary adenomas[J]. Eur Radiol, 2023, 33(5):3312-3321. doi: 10.1007/s00330-023-09412-7 pmid: 36738323
[3]	林绍坚, 吴隽宸, 吴哲褒. 加强垂体神经内分泌肿瘤的免疫治疗研究[J]. 中华内分泌代谢杂志, 2024, 40(11):920-922.
	LIN S J, WU J C, WU Z B, et al. Enhancing resreach on immunotherapy for pitutiary neuroendocrine tumors[J]. Chin J Endocrinol Metab, 2024, 40(11): 920-922.
[4]	LEE D H, PARK J E, NAM Y K, et al. Deep learning-based thin-section MRI reconstruction improves tumour detection and delineation in pre- and post-treatment pitui-tary adenoma[J]. Sci Rep, 2021, 11(1):21302.
[5]	TRITOS N A, MILLER K K. Diagnosis and management of pituitary adenomas: A review[J]. JAMA, 2023, 329(16):1386-1398. doi: 10.1001/jama.2023.5444 pmid: 37097352
[6]	BONNEVILLE J F, POTORAC J, BECKERS A. Neuroimaging of aggressive pituitary tumors[J]. Rev Endocr Metab Disord, 2020, 21(2):235-242.
[7]	KIM M, KIM H S, KIM H J, et al. Thin-slice pituitary MRI with deep learning-based reconstruction: diagnostic performance in a postoperative setting[J]. Radiology, 2021, 298(1):114-122. doi: 10.1148/radiol.2020200723 pmid: 33141001
[8]	KIM M, KIM H S, PARK J E, et al. Thin-slice pituitary MRI with deep learning-based reconstruction for preo-perative prediction of cavernous sinus invasion by pitui-tary adenoma: a prospective study[J]. AJNR Am J Neuroradiol, 2022, 43(2):280-285.
[9]	RASTOGI A, BRUGNARA G, FOLTYN-DUMITRU M, et al. Deep-learning-based reconstruction of under-sampled MRI to reduce scan times: a multicentre, retrospective, cohort study[J]. Lancet Oncol, 2024, 25(3):400-410.
[10]	严福华. 深度学习图像重建算法的临床应用和发展前景[J]. 中华放射学杂志, 2022, 56(11):1165-1167.
	YAN F H. The clinical application and development prospect of deep learning reconstruction algorithm[J]. Chin J Radiol, 2022, 56(11):1165-1167.
[11]	VAN DER VELDE N, HASSING H C, BAKKER B J, et al. Improvement of late gadolinium enhancement image quality using a deep learning-based reconstruction algorithm and its influence on myocardial scar quantification[J]. Eur Radiol, 2021, 31(6):3846-3855. doi: 10.1007/s00330-020-07461-w pmid: 33219845
[12]	DUBLJEVIC N, MOORE S, LAUZON M L, et al. Effect of MR head coil geometry on deep-learning-based MR image reconstruction[J]. Magn Reson Med, 2024, 92(4):1404-1420.
[13]	SHIRAISHI K, NAKAURA T, UETANI H, et al. Deep learning-based reconstruction and 3D hybrid profile order technique for MRCP at 3T: evaluation of image qua-lity and acquisition time[J]. Eur Radiol, 2023, 33(11):7585-7594.
[14]	ICHINOHE F, OYAMA K, YAMADA A, et al. Usefulness of breath-hold fat-suppressed T2-weighted images with deep learning-based reconstruction of the liver: comparison to conventional free-breathing turbo spin echo[J]. Invest Radiol, 2023, 58(6):373-379.
[15]	方姝, 吴梦雄, 陈乾, 等. 深度学习在肝脏屏气T2加权成像图像质量评价中的应用研究[J]. 磁共振成像, 2023, 14(5):31-35,40.
	FANG S, WU M X, CHEN Q, et al. Clinical feasibility of breath-hold fat-suppressed T2-weighted sequence with deep learning reconstruction for liver imaging[J]. Chin J Magn Reson Imaging, 2023, 14(5): 31-35+40.
[16]	CHAIKA M, AFAT S, WESSLING D, et al. Deep learning-based super-resolution gradient echo imaging of the pancreas: Improvement of image quality and reduction of acquisition time[J]. Diagn Interv Imaging, 2023, 104(2):53-59.
[17]	ALMANSOUR H, GASSENMAIER S, NICKEL D, et al. Deep learning-based superresolution reconstruction for upper abdominal magnetic resonance imaging: An analysis of image quality, diagnostic confidence, and lesion conspicuity[J]. Invest Radiol, 2021, 56(8):509-516. doi: 10.1097/RLI.0000000000000769 pmid: 33625063
[18]	SAUER S T, CHRISTNER S A, LOIS A M, et al. Deep learning k-space-to-image reconstruction facilitates high spatial resolution and scan time reduction in diffusion-weighted imaging breast MRI[J]. J Magn Reson Imaging, 2024, 60(3):1190-1200.
[19]	LEE K L, KESSLER D A, DEZONIE S, et al. Assessment of deep learning-based reconstruction on T2-weighted and diffusion-weighted prostate MRI image quality[J]. Eur J Radiol, 2023, 166:111017.
[20]	FEUERRIEGEL G C, WEISS K, KRONTHALER S, et al. Evaluation of a deep learning-based reconstruction method for denoising and image enhancement of shoulder MRI in patients with shoulder pain[J]. Eur Radiol, 2023, 33(7):4875-4884. doi: 10.1007/s00330-023-09472-9 pmid: 36806569
[21]	ESTLER A, HAUSER T K, BRUNNÉE M, et al. Deep learning-accelerated image reconstruction in back pain-MRI imaging: reduction of acquisition time and improvement of image quality[J]. Radiol Med, 2024, 129(3):478-487. doi: 10.1007/s11547-024-01787-x pmid: 38349416
[22]	KANIEWSKA M, DEININGER-CZERMAK E, GETZMANN J M, et al. Application of deep learning-based image reconstruction in MR imaging of the shoulder joint to improve image quality and reduce scan time[J]. Eur Radiol, 2023, 33(3):1513-1525.
[23]	YOO H, YOO R E, CHOI S H, et al. Deep learning-based reconstruction for acceleration of lumbar spine MRI: a prospective comparison with standard MRI[J]. Eur Radiol, 2023, 33(12):8656-8668. doi: 10.1007/s00330-023-09918-0 pmid: 37498386
[24]	XIE Y, LI X, HU Y, et al. Deep learning reconstruction for turbo spin echo to prospectively accelerate ankle MRI: A multi-reader study[J]. Eur J Radiol, 2024, 175:111451.
[25]	SUH P S, PARK J E, ROH Y H, et al. Improving diagnostic performance of MRI for temporal lobe epilepsy with deep learning-based image reconstruction in patients with suspected focal epilepsy[J]. Korean J Radiol, 2024, 25(4):374-383. doi: 10.3348/kjr.2023.0842 pmid: 38528695
[26]	CHAIKA M, BRENDEL J M, URSPRUNG S, et al. Deep learning reconstruction of prospectively accelerated MRI of the pancreas: clinical evaluation of shortened breath-hold examinations with dixon fat suppression[J/OL]. Invest Radiol, 2024-07-23. https://journals.lww.com/investigativeradiology/abstract/9900/deep_learning_reconstruction_of_prospectively.235.aspx.
[27]	SHANBHOGUE K, TONG A, SMEREKA P, et al. Accele-rated single-shot T2-weighted fat-suppressed (FS) MRI of the liver with deep learning-based image reconstruction: qualitative and quantitative comparison of image quality with conventional T2-weighted FS sequence[J]. Eur Radiol, 2021, 31(11):8447-8457.
[28]	PARK H, NAM Y K, KIM H S, et al. Deep learning-based image reconstruction improves radiologic evaluation of pituitary axis and cavernous sinus invasion in pitui-tary adenoma[J]. Eur J Radiol, 2023, 158:110647.
[29]	GONG K, HAN P, EL FAKHRI G, et al. Arterial spin labeling MR image denoising and reconstruction using unsupervised deep learning[J]. NMR Biomed, 2022, 35(4):e4224.
[30]	KAKIGI T, SAKAMOTO R, TAGAWA H, et al. Diagnostic advantage of thin slice 2D MRI and multiplanar reconstruction of the knee joint using deep learning based denoising approach[J]. Sci Rep, 2022, 12(1):10362. doi: 10.1038/s41598-022-14190-1 pmid: 35725760
[31]	DALY A F, BECKERS A. The epidemiology of pituitary adenomas[J]. Endocrinol Metab Clin North Am, 2020, 49(3):347-355.
[32]	LIU Z, WEN B, WANG Z, et al. Deep learning-based reconstruction enhances image quality and improves diagnosis in magnetic resonance imaging of the shoulder joint[J]. Quant Imaging Med Surg, 2024, 14(4):2840-2856.
[33]	WANG X, MA J, BHOSALE P, et al. Novel deep learning-based noise reduction technique for prostate magnetic resonance imaging[J]. Abdom Radiol (NY), 2021, 46(7):3378-3386. doi: 10.1007/s00261-021-02964-6 pmid: 33580348
[34]	TAO Q, WANG K, WEN B, et al. Assessment of image quality and diagnostic accuracy for cervical spondylosis using T2W-STIR sequence with a deep learning-based reconstruction approach[J]. Eur Spine J, 2024, 33(8):2982-2996.

Subjective evaluation indicators	DL T1WI FS			OR T1WI FS
Subjective evaluation indicators	Doctor 1	Doctor 2	ICC	Doctor 1	Doctor 2	ICC
Uniformity	4.33(3,5)	4.35(4,5)	0.92(0.85~0.96)	3.73(3,4)	3.78(3,4)	0.88(0.76~0.93）
Sharpness	4.25(3,5)	4.28(3,5)	0.90(0.81~0.95)	3.50(3,4)	3.53(3,4)	0.86(0.73~0.93)
Artifacts	4.35(4,5)	4.38(4,5)	0.84(0.71~0.91)	2.95(2,4)	2.93(2,4)	0.87(0.75~0.93)
Recognition of pituitary structure	4.38(3,5）	4.45(4,5)	0.93(0.87~0.96)	3.35(2,5)	3.38(3,5)	0.89(0.79~0.94)
Recognition of the lesion	4.6(3,5)	4.53(3,5)	0.85(0.71~0.92)	3.15(2,4)	3.18(2,4)	0.87(0.74~0.93)
Overall quality	4.30(4,5)	4.4(4,5)	0.89(0.79~0.94)	2.63(2,3)	2.68(2,3)	0.88(0.77~0.94)

Objective evaluation indicators	DL T1WI FS			OR T1WI FS
Objective evaluation indicators	Doctor 1	Doctor 2	ICC	Doctor 1	Doctor 2	ICC
SNR _adenoma	26.34(11.42,45.55)	28(10.45，70.66）	0.859(0.73~0.93）	16.76(7.8,30.26）	17.39(8.06,34.55）	0.912(0.83~0.95）
CNR _adenoma	13.72(3.56,30.9)	12.65(0.41,31.81）	0.949(0.90~0.97）	4.44(1.12,8.51）	4.68(1.9,10.49）	0.889(0.79~0.94）
SNR _pituitary	39.92(18.76,67.79)	39.72(21.82,75.51)	0.934(0.88~0.97）	18.42(11.21,28.43）	18.2(11.64,26.41）	0.876(0.77~0.93）
CNR _pituitary	32.12(13.36,63.39)	30.46(10.45,55.09)	0.891(0.79~0.94）	20.73(10.89,42.76）	21.4(11.61,43.8）	0.907(0.83~0.95）

Subjective evaluation indicators	DL T1WI FS	OR T1WI FS	Value of Z	Value of P
Uniformity	4.33(3,5)	3.73(3,4)	-4.69	<0.001
Sharpness	4.25(3,5)	3.50(3,4)	-5.04	<0.001
Artifacts	4.35(4,5)	2.95(2,4)	-7.96	<0.001
Recognition of pituitary structure	4.38(3,5）	3.35(2,5)	-6.14	<0.001
Recognition of the lesion	4.6(3,5)	3.15(2,4)	-7.25	<0.001
Overall quality	4.30(4,5)	2.63(2,3)	-8.04	<0.001

Objective evaluation indicators	DL T1WI FS	OR T1WI FS	Value of Z	Value of P
SNR_adenoma	26.96(18.10,34.15)	16.51(11.24,20.65)	-4.44	<0.000 1
CNR_adenoma	11.30(6.74,19.72)	4.34(3.07,6.00)	-5.55	<0.000 1
SNR _pituitary	38.36(31.93,47.03)	17.02(15.49,20.51)	-5.44	<0.000 1
CNR _pituitary	29.89(23.28,39.75)	18.44(16.61,24.56)	-4.27	<0.000 1