OBJECTIVES: To evaluate the effect of artificial intelligence-assisted compressed sensing (ACS) acceleration on MRI radiomic feature extraction and performance of diagnostic staging models for nasopharyngeal carcinoma (NPC) in comparison with conventional parallel imaging (PI). METHODS: A total of 64 patients with newly diagnosed NPC underwent 3.0T MRI using axial T1-weighted (T1W), T2-weighted (T2W), and contrast-enhanced T1-weighted (CE-T1W) sequences. Both PI and ACS protocols were performed using identical imaging parameters. The total scan time for the 3 sequences in ACS group was 227 s, representing a 30% reduction from 312 s in the PI group. Eighteen first-order and 75 texture features were extracted using Pyradiomics. Intraclass correlation coefficients (ICCs) were calculated to assess the agreement between the two acceleration methods. After feature selection using the least absolute shrinkage and selection operator (LASSO), random forest regression models were constructed to distinguish early-stage (T1 and T2) from advanced-stage (T3 and T4) NPC. The diagnostic performance of the models was evaluated using the area under the receiver operating characteristic curve (AUC) and compared using the DeLong test. RESULTS: ACS-accelerated images demonstrated good radiomic reproducibility, with 86.0% (240/279) of features showing good agreement (ICC>0.75), with mean ICCs for T1W, T2W and CE-T1W sequences of 0.91±0.09, 0.89±0.13 and 0.88±0.11, respectively. The staging prediction models achieved similar AUCs for ACS and PI (0.89 vs 0.90, P=0.991). CONCLUSIONS The MRI radiomic features extracted using ACS and PI techniques are highly consistent, and the ACS-based model shows comparable diagnostic performance to the PI-based model, but ACS significantly reduces the scan time and provides an efficient and reliable acceleration strategy for radiomics in NPC.
Humans ; Nasopharyngeal Neoplasms/diagnosis* ; Magnetic Resonance Imaging/methods* ; Nasopharyngeal Carcinoma ; Neoplasm Staging ; Artificial Intelligence ; Carcinoma ; Female ; Male ; Middle Aged ; Adult ; Radiomics

Purpose To explore the feasibility and clinical value of deep learning-based image reconstruction technology in 5.0T MRI for nasopharyngeal carcinoma.Materials and Methods A prospective study was conducted on 50 newly diagnosed nasopharyngeal carcinoma patients from August to December 2024 at Sun Yat-sen University Cancer Center.5.0T MRI was performed to scan the nasopharynx region.Routine scanning protocols included transverse T2WI,transverse T1WI,transverse contrast-enhanced T1WI and coronal fat-suppressed contrast-enhanced T1WI sequences.Based on these standard scanning protocols,DeepRecon deep learning reconstruction technology with different levels(grade 1-5)was applied,generating a total of 24 sets of images.Qualitative evaluation employed a Likert scale(5-point system)for subjective scoring on lesion detection,lesion edge clarity,artifacts and overall image quality.Quantitative evaluation was performed using the signal-to-noise ratio and contrast-to-noise ratio to objectively assess the quality of the 24 image sets.Differences in qualitative and quantitative indicators between different groups were compared,while the Kappa coefficient was used to analyze the consistency of subjective evaluations by two radiologists.Results In the qualitative assessment of 24 image sets from four MRI sequences(with and without DeepRecon reconstruction),DeepRecon images(grade 2-4)significantly outperformed traditional images in all features except for artifact reduction(Z=-12.11--6.23,all P<0.001).Images reconstructed at DeepRecon grade 3 had the highest overall score and the best image quality.Furthermore,compared with traditional images,DeepRecon images(grade 2-5)demonstrated significantly improved signal-to-noise ratio for both lesions and the lateral pterygoid muscle(t=-15.67--3.44,Z=-6.09--4.63,all P<0.01).In addition,in the transverse T2WI,transverse contrast-enhanced T1WI and coronal fat-suppressed contrast-enhanced T1WI images with DeepRecon reconstruction(grade 2-5),the contrast-to-noise ratio(lesion/lateral pterygoid muscle)also showed significant improvement compared to traditional images(t=-12.71--3.19,Z=-6.08--4.47,all P<0.001).The inter-observer agreement for the overall subjective quality score between the two radiologists was good(Kappa=0.75-0.82,all P<0.01).Conclusion DeepRecon deep learning reconstruction technology significantly increases the signal-to-noise ratio and resolution of traditional magnetic resonance images of nasopharyngeal cancer,improving image clarity and bringing more possibilities for the advancement of imaging diagnosis.

Purpose To explore the feasibility and clinical value of deep learning-based image reconstruction technology in 5.0T MRI for nasopharyngeal carcinoma.Materials and Methods A prospective study was conducted on 50 newly diagnosed nasopharyngeal carcinoma patients from August to December 2024 at Sun Yat-sen University Cancer Center.5.0T MRI was performed to scan the nasopharynx region.Routine scanning protocols included transverse T2WI,transverse T1WI,transverse contrast-enhanced T1WI and coronal fat-suppressed contrast-enhanced T1WI sequences.Based on these standard scanning protocols,DeepRecon deep learning reconstruction technology with different levels(grade 1-5)was applied,generating a total of 24 sets of images.Qualitative evaluation employed a Likert scale(5-point system)for subjective scoring on lesion detection,lesion edge clarity,artifacts and overall image quality.Quantitative evaluation was performed using the signal-to-noise ratio and contrast-to-noise ratio to objectively assess the quality of the 24 image sets.Differences in qualitative and quantitative indicators between different groups were compared,while the Kappa coefficient was used to analyze the consistency of subjective evaluations by two radiologists.Results In the qualitative assessment of 24 image sets from four MRI sequences(with and without DeepRecon reconstruction),DeepRecon images(grade 2-4)significantly outperformed traditional images in all features except for artifact reduction(Z=-12.11--6.23,all P<0.001).Images reconstructed at DeepRecon grade 3 had the highest overall score and the best image quality.Furthermore,compared with traditional images,DeepRecon images(grade 2-5)demonstrated significantly improved signal-to-noise ratio for both lesions and the lateral pterygoid muscle(t=-15.67--3.44,Z=-6.09--4.63,all P<0.01).In addition,in the transverse T2WI,transverse contrast-enhanced T1WI and coronal fat-suppressed contrast-enhanced T1WI images with DeepRecon reconstruction(grade 2-5),the contrast-to-noise ratio(lesion/lateral pterygoid muscle)also showed significant improvement compared to traditional images(t=-12.71--3.19,Z=-6.08--4.47,all P<0.001).The inter-observer agreement for the overall subjective quality score between the two radiologists was good(Kappa=0.75-0.82,all P<0.01).Conclusion DeepRecon deep learning reconstruction technology significantly increases the signal-to-noise ratio and resolution of traditional magnetic resonance images of nasopharyngeal cancer,improving image clarity and bringing more possibilities for the advancement of imaging diagnosis.

Objective To explore the diagnostic value of MRE for differentiating hepatic benign and malignant tumors.Methods 36 patients with liver tumor (a total of 39 lesions,including 20 hepatocellular carcinomas,7 hemangiomas,5 cholangiocellular carcinomas,3 metastases,2 hepatic angiomyolipomas,1 carcinosarcoma,1 castleman’s disease)and 9 healthy volunteers were evaluated with MRE.The elastogram were generated with FUNCTOOL post processing program.The mean value of elasticity of hepatic malignant tumors,hepatic benign tumors,hepatic parenchyma around the malignant tumors,hepatic parenchyma around the benign tumors and the normal liver of healthy volunteers were measured and compared.Results The mean value of elasticity of malignant tumors [(7.39±1.70)kPa]was significantly higher than these of benign tumors [(4.11±0.37)kPa,P < 0.001],peripheral parenchyma around the malignant tumors [(3.50±0.73)kPa,P < 0.001],peripheral parenchyma around the benign tumors [(2.61±0.45)kPa,P < 0.001] and normal liver of healthy volunteers [(2.38±0.24)kPa,P <0.001].The mean value of elasticity of parenchyma around the malignant tumors [(3.50±0.73)kPa]was significantly higher than that of hepatic parenchyma around the benign tumors [(2.61 ± 0.45)kPa,P <0.001].The mean value of elasticity of hepatic parenchyma around the benign tumors [(2.61±0.45)kPa]was slightly higher than normal liver of healthy volunteers [(2.38±0.24)kPa],and there was no significant difference between the two (P >0.05).A cutoff value of 5.08 kPa can accurately differentiate malignant tumors from benign tumors and normal liver parenchyma.Conclusion MRE could be used in diagnosis of hepatic focal tumors,which is helpful for differentiating benign and malignant liver tumors.