- Volume 20, Issue 1, 2024

Meniscal tears are among the most common indications for knee arthroscopy. Artificial polyurethane scaffolds have demonstrated efficacy in reducing pain and promoting the growth of normal meniscal tissue, with high absorption rates facilitating full tissue regeneration.

This study aimed to evaluate the remodeling of polyurethane meniscal implants post-reconstruction using ultrasonography. This imaging technique not only assesses changes in implant properties, such as echogenicity, but also the shape changes during functional examination.

The assessment of meniscal extrusion, comparing size at rest and under weight-bearing, is an indirect parameter that provides insight into the physical properties of the remodeling implant, with greater extrusion indicating reduced stiffness and inferior physical properties of the meniscus. Ultrasonography has the valuable advantage of allowing for assessment of the blood supply to the meniscus through Power Doppler imaging.

The presence of vessels within the meniscal implants serves as evidence of ongoing remodeling. The study included 35 patients (13 female, 22 male; mean age 41.6 years, range 18-66) who underwent arthroscopic meniscal reconstruction with polyurethane implants, with an average time from surgery of 2.8 years (range 0.3-4.5 years). Results showed complete (29.7%), significant (45.9%), or moderate (16.2%) remodeling into natural meniscal tissue in 91.8% of the implants.

The mean values of extrusion in the supine position and during 90-degree flexion were significantly greater in the operated limb (2.603) compared to the contralateral limb (1.978; t(35) = 2.442; P < 0.05). No significant differences in extrusion were found between the limbs in a standing position, indicating favorable physical properties of the polyurethane meniscal implants. Further ultrasonography studies of meniscal scaffolds are deemed relevant.

Rheumatoid Arthritis Magnetic Resonance Imaging Score (RAMRIS) is usually used for the semi-quantitative evaluation of joint changes in Rheumatoid Arthritis (RA). However, this method cannot evaluate early changes in bone marrow edema (BME).

To determine whether T1 mapping of wrist BME predicts early treatment response in RA.

This study prospectively enrolled 48 RA patients administered oral anti-rheumatic drugs. MRI of the most severely affected wrist was performed before and after 4 (48 patients) and 8 weeks of treatment (38 patients). Mean T1 values of BME in the lunate, triangular, and capitate bones; RAMRIS for each wrist; Erythrocyte-Sedimentation Rate (ESR); and 28-joint Disease Activity Score (DAS28)-ESR score were analyzed. Patients were divided into responders (4 weeks, 30 patients; 8 weeks, 32 patients) and non-responders (4 weeks, 18 patients; 8 weeks, 6 patients), according to EULAR response criteria. Receiver operating characteristic (ROC) curves were used to evaluate the efficacy of T1 values.

ESR and DAS28-ESR were not correlated with T1 value and RAMRIS at each examination (P > 0.05). Changes in T1 value and DAS28-ESR relative to the baseline were moderately positively correlated with each other at 4 and 8 weeks (r = 0.555 and 0.527, respectively; P < 0.05). At 4 weeks, the change and rate of change in T1 value significantly differed between responders and non-responders (-85.63 vs. -19.92 ms; -12.89% vs. -2.81%; P < 0.05). The optimal threshold of the rate of change in T1 value at 4 weeks for predicting treatment response was -5.32% (area under the ROC curve, 0.833; sensitivity, 0.900; specificity, 0.667).

T1 mapping provides a new imaging method for monitoring RA lesions; changes in wrist BME T1 values reflect early treatment response.

This study sought to analyze the ¹⁸F-FDG PET/CT and contrast-enhanced computed tomography (CT) images of synchronous colorectal cancer (CRC) and renal clear cell carcinoma (ccRCC) and identify the shared genes between these two types of cancer through bioinformatic analysis.

A retrospective analysis was conducted on a patient with synchronous CRC and ccRCC who underwent ¹⁸F-FDG PET/CT and contrast-enhanced CT before treatment. Databases were analyzed to identify differentially expressed genes between CRC and ccRCC, and co-expression genes were extracted for RCC and CRC.

¹⁸F-FDG PET/CT revealed intense metabolic activity in the primary colorectal lesion (SUVmax 13.2), while a left renal mass (diameter = 35 mm) was observed with no significant uptake. Contrast-enhanced CT during the arterial phase showed heterogeneous intense enhancement of the renal lesion, and the lesion washed out earlier than in the renal cortex in the nephrographic and excretory phases, indicating ccRCC. The histopathological results confirmed synchronous double primary malignant tumors. Our bioinformatic analysis results showed that synchronous occurrence of CRC and ccRCC may correlate with simultaneous expression of Carbonic Anhydrase 9 (CA9), integrin-binding sialoprotein (IBSP), and Fibrinogen γ chain (FGG).

¹⁸F-FDG PET/CT combined with contrast-enhanced CT is an effective diagnostic tool in evaluating synchronous CRC and RCC. By analyzing this clinical case and conducting bioinformatic analysis, we improved our current understanding of the mechanisms underlying synchronous tumors.

Generative adversarial networks (GANs) have demonstrated superior data generation capabilities compared to other methods, making them popular for use in medical image applications. These features have intrigued researchers in the medical imaging field, resulting in a swift implementation of these techniques in various conventional and novel applications such as image reconstruction, segmentation, detection, classification, and cross-modality synthesis. A comprehensive review of recent medical imaging breakthroughs will benefit researchers interested in this field. In this review, we aimed to introduce the origin, principle, and extended forms of GANs and summarize the state-of-the-art progress of GAN-based medical image processing methods.

We searched the literature for studies on Google Scholar and PubMed using the keywords “Segmentation,” “Classification,” “medical image,” and “generative adversarial network.” Specifically, the initial search revealed 5423 publications after the removal of duplicated and non-accessible full-text publications. Then, after the title and abstract screening, 680 underwent full-text screening. Finally, 121 studies were included in our final analysis after full-text screening.

The date range of the studies covered in this review is from January 1, 2017, to the present. After a thorough screening and qualification assessment, 121 studies involving GAN-based applications in seven areas of medical images were included in the final methodological review. These areas included synthesis, classification, segmentation, conversion, reconstruction, denoising, and lesion detection. We further classified and summarized these papers into clinical applications, classification methods, and imaging modalities.

We thoroughly examined the latest research progress of GAN-based medical image augmentation. These techniques effectively alleviate the challenge of limited training samples for medical image diagnosis and treatment models. Furthermore, several critical issues associated with GANs, such as pattern collapse, instability, and lack of interpretability, require attention in future research.

The use of breast MRI for screening has increased over the past decade, mostly in women with a high risk of breast cancer. Abbreviated breast MRI (AB-MR) is introduced to make MRI a more accessible screening modality. AB-MR decreases scanning and reporting time and the overall cost of MRI.

This study aims to evaluate the diagnostic efficacy of abbreviated MRI protocol in detecting breast cancer in screening and diagnostic populations, using histopathology as the reference standard.

This is a single-centre retrospective cross-sectional study of 134 patients with 198 histologically proven breast lesions who underwent full diagnostic protocol contrast-enhanced breast MRI (FDP-MR) at the University Malaya Medical Centre (UMMC) from 1st January 2018 to 31st December 2019. AB-MR was pre-determined and evaluated with regard to the potential to detect and exclude malignancy from 3 readers of varying radiological experiences. The sensitivity of both AB-MR and FDP-MR were compared using the McNemar test, where both protocols' diagnostic performances were assessed via the receiver operating characteristic (ROC) curve. Inter-observer agreement was analysed using Fleiss Kappa.

There were 134 patients with 198 lesions. The average age was 50.9 years old (range 27 – 80). A total of 121 (90%) MRIs were performed for diagnostic purposes. Screening accounted for 9.4% of the cases, 55.6% (n=110) lesions were benign, and 44.4% (n=88) were malignant. The commonest benign and malignant lesions were fibrocystic change (27.3%) and invasive ductal carcinoma (78.4%). The mean sensitivity, specificity, positive predictive value, and negative predictive value for AB-MR were 0.96, 0.57, 0.68 and 0.94, respectively. Both AB-MR and FDP-MR showed excellent diagnostic performance with AUC of 0.88 and 0.96, respectively. The general inter-observer agreement of all three readers for AB-MR was substantial (k=0.69), with fair agreement demonstrated between AB-MR and FDP-MR (k=0.36).

The study shows no evidence that the diagnostic efficacy of AB-MR is inferior to FDP-MR. AB-MR, with high sensitivity, has proven its capability in cancer detection and exclusion, especially for biologically aggressive cancers.

Breast cancer is one of the leading causes of mortality among women. In addition, 1 in 8 women and 1 in 833 men will be diagnosed with breast cancer in 2022. The detection of breast cancer can not only lower treatment costs but also increase survival rates. Due to increased cancer awareness, more women are undergoing breast cancer screening, leading to more cases being diagnosed worldwide, but doctors' ability to analyze these images is limited. As a result, they get overloaded leading to misinterpretations. The advent of computer-aided diagnosis (CAD) minimized man’s involvement and achieved good results. CAD helps medical doctors automatically detect and analyze abnormalities found in the breast. Such abnormalities may be benign or malignant tumors.

The goal of this study is to evaluate the effectiveness of using seven layers to classify breast cancer as either benign or malignant using mammograms.

The open-source MIAS dataset of 322 images was used for our study, of which 207 were normal images and 115 were abnormal images. The proposed CNN model convolves an image into seven layers that extract features from the input images, and these features are used to classify breast cancer as malignant or benign.

The proposed CNN used a limited data set and achieved the best result compared to previous work. The method achieved results with a 0.39% loss, 99.89% accuracy, 99.85% precision, 99.89% recall, 99.87% F1-score, and an area under the curve noted to be 100.0%.

CNN uses a small amount of data to determine abnormalities; the method will assist a medical doctor in determining whether or not a specific patient has cancer.

Ultrasound is extensively used for soft tissue pathology. Scanning bone superficial structures may reveal clear pathologic features to aid diagnosis.

We present the case of a stress fracture in the second metatarsal, with the clinical aspect of a gouty attack. Ultrasound examination showed cortical thickening and disruption, hypoechoic periosteal swelling, hyperemia, soft tissue edema, and displacement of the extensor tendon. The diagnosis was confirmed by X-ray and MRI. The value of different diagnostic tools is discussed, and the importance of gray-scale and color Doppler ultrasound as a first-hand modality is underlined.

Sonography clearly identifies cortical and periosteal abnormalities, differentiates surrounding soft tissue pathologies, and offers dynamic evaluation, and follow-up possibility with low cost, high accessibility, and no risks. Periosteal and cortical irregularities are important diagnostic issues when performing ultrasound examinations for soft tissue pathology.

Liposarcoma (LPS) is a malignant mesenchymal tumor that mostly occurs in the extremities and retroperitoneum and rarely in the scrotum.

In this case report, we introduced a patient who was diagnosed with LPS in the scrotum. In his right scrotum, we found a large soft tissue mass, including fat and calcification.

We reviewed the clinical, pathological and computed tomography (CT) features of patients diagnosed with LPS of the scrotum to help improve the understanding of the disease and the accuracy of diagnosis.

This work aimed to evaluate the level set segmentation algorithm on ocular surface thermograms. In addition, the vascularity functioning between the contralateral portions of two eyes (right and left) was identified using statistical analysis methods.

A total of 25 healthy participants with an average age of 35 years (20 men and 5 women) were selected in April 2022. Thermogram images were captured using a FLIR T series thermal camera. Conventional image processing techniques, such as filtering and edge detection, were used to preprocess thermograms. Next, the level set approach was used with the edge-detected pattern as an input to an automated segmented region of interest (ROI).

Five metrics, namely Dice Coefficient, Tanimoto Index, Jaccard Index, Volume Similarity, and Structural Similarity, were used to assess the performance of the segmentation technique compared to ground truth, which showed 97.5%, 92.5%, 94.5%, 96.5%, and 96.5% correlation, respectively, between the segmented and the ground truth images with average values for both the eyes. Statistical analysis demonstrated that the contralateral portions of the ocular thermograms were significantly different in terms of vascular distribution between the left and right eyes (p < 0.005).

The level set method efficiently segmented the ROI in ocular thermograms with maximum correlation. According to the segmentation’s results, the model showed the dissimilarity between the contralateral parts of the left and right eyes in healthy cases.

This study aims to investigate the efficiency of a radiomics model in identifying high-frequency microsatellite instability (MSI-H) and microsatellite stability (MSS) of colorectal liver metastasis (CRLM) according to machine learning radiomics features of enhanced CT liver images.

A total of 12 patients with MSI-H CRLM and 96 patients with MSS CRLM were randomly divided into the training group and internal validation group according to the ratio of 7: 3 (training: 75 cases, validation: 33 cases). From the enhanced CT (portal phase) image data of patients, 788 radiomics features were extracted, and a random forest model was established with the optimal features selected. The receiver operating characteristics (ROC) curve analysis was performed to assess the model’s diagnostic efficacy.

The training group comprised 8 patients with MSI-H CRLM and 67 patients with MSS CRLM, and the internal validation group included 4 patients with MSI-H CRLM and 29 patients with MSS CRLM. After feature selection, 7 radiomics features good for distinguishing MSI-H CRLM and MSS CRLM were screened out. The ROC curve analysis demonstrated that the random forest model had the AUC (area under the ROC curve) value 0.88, accuracy 0.85, sensitivity 0.85, specificity 0.92, and F1 score 0.88 in the training group. The model had an AUC value of 0.75, accuracy of 0.74, sensitivity of 0.81, specificity of 0.85, and F1_score of 0.78 in the internal validation group in identifying the MSI-H from the MSS CRLM. In order to evaluate the robustness of the overall model, the 788 features obtained were all applied to the 5-fold cross-validation, with the model being built on the random forest and analyzed with the ROC curve analysis. The AUC value of the model was 0.86 (P<0.05), accuracy value 0.91, sensitivity 0.60, and specificity 0.95.

The random forest prediction model built on the radiometric features extracted from enhanced CT images can be used to identify the MSI-H from the MSS CRLM and may provide effective guidance for clinical immunotherapy of CRLM patients with unknown MSI status.

Betel nut chewing is very common in Southeast Asia and other tropical countries. Much clinical evidence suggests that chewing betel nut has pro-inflammatory and carcinogenic effects, but there are few clinical reports of acute toxicity caused by it, especially involving esophageal damage.

We presented a case of a 72-year-old female who was admitted to our hospital for chest pain and hematemesis within several minutes after chewing betel nut. Gastroscopy showed two longitudinal ridge-like mucosal eminences in the esophagus located 20 cm from the incisors down to the gastric cardia, which was similar to varices. At last, a CT scan showed concentric-circle thickening of the esophagus wall, suggesting hematomas. Our treatment included fasting, inhibiting gastric acid and maintaining blood volume. After one week of medical treatment, rechecked gastroscopy showed that esophageal hematomas were gradually absorbed, with the formation of multiple shallow ulcers.

The acute toxicity of chewing betel nut can be easily overlooked. Patients who experience chest pain or hematemesis after chewing betel nut products,especially those who take aspirin at the same time, need to be alert to esophageal hematoma.

Millions of people have been infected with COVID-19, which has spread quickly worldwide since the start of 2020, resulting in numerous fatalities. Identification of infected individuals is essential to control the spread of the virus.

In this study, we propose a hybrid architecture that combines Convolutional Neural Networks (CNNs) with Recurrent Neural Networks (RNNs) and leverages transfer learning to enhance the accuracy of COVID-19 detection from X-ray images.

The proposed work utilizes 4 pre-trained CNN architectures, namely, InceptionnetV3, Densenet121, Inception-ResNet V2, and VGG19, to extract high-level features from the input X-ray images. These features are then fed into the second component, an RNN-based network, which captures the temporal dependencies within the extracted features. To evaluate the performance of the proposed architecture, a comprehensive dataset consisting of X-ray images from COVID-19 positive cases, non-COVID-19 pneumonia cases, and healthy individuals is used. Gradient class activation map (Grad-CAM) analysis has been applied to the obtained results to provide heat-map pictures specific to each class and coloured visualizations of the COVID-19-infected areas in CXR images.

: Experimental results demonstrate that the proposed hybrid CNN-RNN architecture achieves promising results in COVID-19 detection from X-ray images. The model exhibits high accuracy, precision, recall, area under the receiver operating characteristics (ROC) curve (AUC), and F1-score, outperforming other state-of-the-art methods.

The combination of CNNs and RNNs enables the model to effectively capture spatial and temporal information, leading to improved performance in COVID-19 detection. The proposed hybrid architecture with transfer learning from X-ray images provides a robust and efficient solution for COVID-19 detection. The model can potentially assist healthcare professionals in making accurate and timely diagnoses, thereby contributing to the global efforts to combat the COVID-19 pandemic. In the present work, VGG19-RNN architecture outperformed all other networks in terms of accuracy. The most effective training and validation accuracy for the VGG19-RNN architecture is 99% & 97.70%, respectively, and the loss was 0.02 & 0.09 at epoch 100.

Brain tumors are predicted from Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) scan images. In recent years, image processing-based automated tools are developed to predict tumor areas with less human interference. However, such automated tools are suffering from computational complexity and reduced accuracy in certain critical images. In the proposed work, an Ideal Shallow Neural Network (ISNN) is utilized to improve the prediction accuracy, and the computational complexity is reduced by implementing an Artificial Jellyfish Optimization (AJO) algorithm for minimizing the feature dimensionality.

The proposed method utilizes MRI images for the verification process as they are more informative than the CT scan image. The BRATS and the Kaggle datasets are used in this work and a Gabor filtering technique is used for noise reduction and a histogram equalization is used for enhancing the tumor boundary regions. The classification results observed from the AJO-ISNN are further forwarded towards the segmentation process and which uses the Centroid Weighted Segmentation (WCS) along with a Grasshopper Optimization Algorithm (GOA) for improving the segmentation over the boundary regions of the brain tumor.

The experimental result indicates a classification accuracy of 95.14% on the proposed AJO-ISNN model and AJO-ISNN is comparatively better than the Convolutional Neural Network (CNN) model accuracy of 85.41% and VGG 19 model accuracy of 93.75% while implemented with the AJO optimization model. Similarly, the Dice Similarity Coefficient of the proposed CWS-GOA also reaches 93.15% when performed with both BRATS and Kaggle datasets.

Apart from the accuracy attainments the proposed work classifies and segments the tumor region in around 65 seconds on average of 200 image verifications and that is comparatively better than the previous multi-cascaded CNN and the InceptionV3 models.

We aimed to examine all-cause mortality risk in relation to the extent of non-obstructive coronary artery disease (CAD) by coronary computed tomography angiography (CTA) in Chinese middle-aged and older patients in a multicenter study with nine-year follow-up.

This was a retrospective, observational, multicentre study. The study population consisted of 3,240 consecutive middle-aged and older patients (age ≥ 40 years) with suspected CAD who underwent coronary CTA between June 2011 and December 2013 at three hospitals in Wuhan, China. Patients were grouped according to CAD extent for the final analysis: no CAD, 1-vessel non-obstructive CAD, 2-vessels non-obstructive CAD, and 3-vessels non-obstructive CAD. The primary endpoint was all-cause mortality. Kaplan-Meier method and Cox proportional hazards regression models were used for analysis.

A total of 2,522 patients were included in the present analysis. Of these, 188 (7.5%) deaths occurred during the median 9.0 years (interquartile range 8.6–9.4) of study follow-up. The annualized all-cause mortality rate was 0.54 (95% CI: 0.44–0.68), 0.91 (95% CI: 0.68–1.21), 1.44 (95% CI: 1.01–1.93), and 2.00 (95% CI: 1.46–2.69) for the no CAD, 1-vessel non-obstructive CAD, 2-vessels non-obstructive CAD, and 3-vessels non-obstructive CAD group, respectively. Kaplan–Meier survival curves showed a significant increase in the cumulative events associated with the extent of non-obstructive CAD (P < 0.001). In multivariate Cox regression, after adjustment for age and sex, the presence of 3-vessels non-obstructive CAD was a significant predictor of all-cause mortality (HR 1.60, 95% CI: 1.04–2.45, P = 0.032).

In this cohort of Chinese middle-aged and older patients undergoing coronary CTA, the presence and extent of non-obstructive CAD, compared to no CAD, were associated with a significantly greater nine-year risk of all-cause mortality. The present findings suggest the clinical importance of the stage of non-obstructive CAD and warrant investigation of the optimal risk stratification to improve outcomes among these patients.

Deep learning-based diagnosis systems are useful to identify abnormalities in medical images with the greatly increased workload of doctors. Specifically, the rate of new cases and deaths from malignancies is rising for liver diseases. Early detection of liver lesions plays an extremely important role in effective treatment and gives a higher chance of survival for patients. Therefore, automatic detection and classification of common liver lesions are essential for doctors. In fact, radiologists mainly rely on Hounsfield Units to locate liver lesions but previous studies often pay little attention to this factor.

In this paper, we propose an improved method for the automatic classification of common liver lesions based on deep learning techniques and the variation of Hounsfield Unit densities on CT images with and without contrast. Hounsfield Unit is used to locate liver lesions accurately and support data labeling for classification. We construct a multi-phase classification model developed on the deep neural networks of Faster R-CNN, R-FCN, SSD, and Mask R-CNN with the transfer learning approach.

The experiments are conducted on six scenarios with multi-phase CT images of common liver lesions. Experimental results show that the proposed method improves the detection and classification of liver lesions compared with recent methods because its accuracy achieves up to 97.4%.

The proposed models are very useful to assist doctors in the automatic segmentation and classification of liver lesions to solve the problem of depending on the clinician’s experience in the diagnosis and treatment of liver lesions.

This article aims to explore the ADC value of ultrahigh b-value DWI and the diagnostic cutoff point in prostate cancer.

A total of 78 patients were included in this study. T2 weighted imaging (T2WI), conventional diffusion-weighted imaging (DWI) (1000 s/mm²), and DWI with ultrahigh b-values of 2000 s/mm² and 3000 s/mm² were performed in each patient. With reference biopsy as the gold standard, the apparent diffusion coefficient (ADC)s of each b-value DWI image were analyzed. According to different b-value receiver operating characteristic (ROC) curves, the ADC diagnostic cutoff point for prostate cancer was determined.

A total of 154 lesions were identified as prostate cancer. The ADC values for conventional DWI and ultrahigh b-value DWI with 2000 s/mm² and 3000 s/mm² were 1.097×10^-3 mm²/s (1.040-1.153), 0.809×10^-3 mm²/s (0.766-0.851) and 0.622×10^-3 mm²/s (0.591-0.652), respectively, in the peripheral zone and 1.085×10^-3 mm²/s (1.022-1.147), 0.815×10^-3 mm²/s (0.770-0.861) and 0.651×10^-3 mm²/s (0.617-0.685) in the transition zone. The area under the curve (AUC)s of the ADC values from ultrahigh b-value DWI (2000 s/mm² and 3000 s/mm²) were 0.824 and 0.852 in the peripheral zone and 0.905 for the ADC values from ultrahigh b-value DWI (3000 s/mm²) in the transition zone. In the peripheral zone, the ADC diagnostic cutoff values for prostate cancer were 0.75×10^-3 mm²/s and 0.685×10^-3 mm²/s in DWI at 2000 s/mm² and 3000 s/mm², respectively, and the diagnosis of transition zone cancer was 0.8×10^-3 mm²/s and 0.634×10^-3 mm²/s, respectively.

The ADC values from ultrahigh b-value DWI demonstrated better consistency and diagnostic efficacy in the diagnosis of prostate cancer.

DWI and ADC-mapping was performed to analyze hepatic metastasis of GIST, GEP-NET.

The objective of this study is to present hepatic metastasis of GIST and GEP-NET with Diffusion weighted MR imaging(DWI) and the Apparent diffusion coefficients (ADC) values of masses.

18 GIST patients and 8 GEP-NET patients were examined retrospectively. 11 males and 6 females were present in GIST group, 7 males to 5 females were involved in GEP-NET group. 18 primary GIST and 10 hepatic metastasis of GIST, 8 original GEP-NET and 19 hepatic metastasis of GEP-NET; total 55 GIST and GEP-NET masses were analysed by ADC mapping. MR images were acquired by 1,5 T MR units (32 mT/min gradient strength- Achieva; Philips Healthcare, Best, Netherlands and 32 channel GE Signa GE-Wisconsin-USA); by using a 4-8 channel standard phased-array torso XL coil, all images were evaluated by an Abdominal MRI experienced radiologist. DWI was performed in the transverse plane by using spin-echo-planar imaging sequence.

No statistical differences were observed between GIST and GEP-NET patients according to age and gender variations. No significant statistical differences were observed according to the diameters and ADC values of GIST and GEP-NET patients. A significant statistical difference was observed between GIST and GEP-NET groups in terms of size of liver metastasis which was significantly higher in GIST patients. All three groups (GIST_Hep. MET, GEP-NET_Liver_Met and normal) were statistically differed according to ADC values. With the ROC curve analysis: Hepatic metastasis of GIST(n=10) and normal liver (n:47) had cut-off value for ADC: 0.925 under AUC: 0.939 with regard to ADC values and regarded 89.4% Sensitivity, 100% Specificity, 100% PPV and 66.7% PPV. ROC curve of GEP NET_ Hepatic metastasis (n=19) group and normal liver (n:47) group presented cut-off value for ADC: 0.860 under AUC: 0.967 correlated to ADC values with 93.6% sensitivity, 89.5% specificity, 95.7% PPV and 85% PPV.

High cellular tumors resulted from liver metastasis of GIST and GEP-NET’s, and a positive correlation was observed between ADC values and cellularity/differentiation ratios of metastatic masses.