Synopsis

Clinical feasibility study of AI accelerateed STAGE

AI SNR

Synopsis:In recent years, a rapid imaging technique named Strategically Acquired Gradient Echo（STAGE）has been actively introduced in clinical practice and scientific research. This technique uses a multi-flip angle and multi-echo acquisition method to create more than ten images, including multiple qualitative and quantitative images, within 9 mins on 1.5T. By AI acceleration, the scan time of this technique is further shortened to about 4 mins.¹ In this work, we have presented a clinical comparison study between routine STAGE and AI accelerated STAGE.This collection method with AI could complete the acquisition at a larger multiple of acceleration a word used in a person's name. Does not affect the image quality. Furthermore, the acceleration has further increased the scope of clinical application. The comparative analysis was performed and accelerate the application of this technology in clinical practiceMethods: 10 patients were enrolled using 1.5T system (NeuMR, Neusoft Medical) from 2022.10.01 to 2022.10.31. Besides, the BMI index, gender and age were not restricted. Patients were scanned undifferentiated using Brain Quant (routine STAGE) and Smart Brain Quant (AI STAGE) using a 24 Head and neck combination (NV) coil. Brain Quant sequence parameters are: TR=45ms, TE=10/23/39ms, matrix size=384*384*48, slice thickness= 2.7mm, average times =1, flip angle= 8,overall acceleration factor=2. Smart BrainQuant scanning parameters were same as Brain Quant except that overall acceleration factor=3.9.The two methods obtained PDWI, T1WI, T2 * WI, eT1WI, SWI, SWIP, tSWI, PDMAP, T1MAP, R2 * MAP, QSM images. Both quantitative and qualitative evaluations were performed. Quantitative measures were used the SNR and CNR matrics. Images were acquired for background and tissue ROI to obtain the SI tissue and SD background. Then through SNR=SI tissue / SD background, SI tissue is the average of the region of interest, and SD background is the standard deviation of background noise. The detection method is a region of interest outside the tissue in the field in the direction of image phase coding, and SD is the standard deviation of the region of interest (You can also use the picture below). CNR The calculation method is Inash White matter Area Painting ROI, pass through mean and variance of pixels within the corresponding region were calculated separately. See formula for details.Quantitative Image Analysis²: Region of Interes (ROI)-Based The iterative reconstruction used for Compressed SENSE leads to an artifcial reduction of noise within the image. This makes the classic measurement approaches for SNR and CNR problematic. In particu- lar, small intensity peaks within the air reflecting noise are reduced by the algorithm and thus background signal cannot be reliably used for the calculation. To still be able to quantify potential differences between the sequences³, we calculated the SNR by dividing the signal intensity(Utisue) within an ROI in the central slice through the stan- dard deviation(Gtisuewithin the same ROI. The CNR was calculated as the difference between the signal intensity of the different tissues divided by the standard devition, as described previously.21 The ROI areas were 150 mm²for bone marrow and 25 mm² for CSF,the nerve roots(level L4/5 on the right side)and spinal cord. Qualitative evaluation method: At the same time, two doctors of the same year were invited to double-blind score the image quality. The score standard was 1 point cannot meet the diagnosis, 2 points meet the diagnosis but poor quality, 3 points meet the diagnostic quality, and 4 points meet the excellent diagnostic quality.Result: Through 10 sets of images, the same level and the same coordinates of the two sequences organization. The ROI performed the measurements and found that the Brain Quant sequences in the phase acceleration factor increases to 2.8 and the level acceleration factor to 1.4 times. The SNR was significantly lower than the Smart Brain Quant with AI support. Measurements of values in T1(AI) SNR11.8-12.6. The original sequence was measured as T1SN using the same acceleration factor R 7.9-10.4. Owing to PD likeness By the B1 field image is larger, so temporarily do not do a reference. The AI group was rated by the rating of 3 points. The original sequence score of 3.2 Points which was fully meet the diagnosis. The quantitative analysis is shown in Table 1.The comparison image were shown in Figure 1Conclusion: Pass through AI blessing under the Smart Brain Quant sequence, in the after expanding the phase acceleration factor and the level acceleration factors. Both the quantitative and qualitative evaluations were good expression, greatly reduced scan time, change the trial, clinical and scientific research situation, improve the patient tolerance and decrease the time cost, which has a good application prospect. Figure1. Image comparation between Smart BrainQuant and BrainQuant Table 1.Sequences mean value

Acknowledgements

References