Modality-based attention and dual-stream multiple instance convolutional neural network for predicting microvascular invasion of hepatocellular carcinoma

Background and purpose

The presence of microvascular invasion (MVI) is a crucial indicator of postoperative recurrence in patients with hepatocellular carcinoma (HCC). Detecting MVI before surgery can improve personalized surgical planning and enhance patient survival. However, existing automatic diagnosis methods for MVI have certain limitations. Some methods only analyze information from a single slice and overlook the context of the entire lesion, while others require high computational resources to process the entire tumor with a three-dimension (3D) convolutional neural network (CNN), which could be challenging to train. To address these limitations, this paper proposes a modality-based attention and dual-stream multiple instance learning(MIL) CNN.

Materials and methods

In this retrospective study, 283 patients with histologically confirmed HCC who underwent surgical resection between April 2017 and September 2019 were included. Five magnetic resonance (MR) modalities including T2-weighted, arterial phase, venous phase, delay phase and apparent diffusion coefficient images were used in image acquisition of each patient. Firstly, Each two-dimension (2D) slice of HCC magnetic resonance image (MRI) was converted into an instance embedding. Secondly, modality attention module was designed to emulates the decision-making process of doctors and helped the model to focus on the important MRI sequences. Thirdly, instance embeddings of 3D scans were aggregated into a bag embedding by a dual-stream MIL aggregator, in which the critical slices were given greater consideration. The dataset was split into a training set and a testing set in a 4:1 ratio, and model performance was evaluated using five-fold cross-validation.

Results

Using the proposed method, the prediction of MVI achieved an accuracy of 76.43% and an AUC of 74.22%, significantly surpassing the performance of the baseline methods.

Conclusion

Our modality-based attention and dual-stream MIL CNN can achieve outstanding results for MVI prediction.

Spectral computed tomography with inorganic nanomaterials: State-of-the-art

Recently, spectral computed tomography (CT) technology has received great interest in the field of radiology. Spectral CT imaging utilizes the distinct, energy-dependent X-ray absorption properties of substances in order to provide additional imaging information. Dual-energy CT and multi-energy CT (Spectral CT) are capable of constructing monochromatic energy images, material separation images, energy spectrum curves, constructing effective atomic number maps, and more. However, poor contrast, due to neighboring X-ray attenuation of organs and tissues, is still a challenge to spectral CT. Hence, contrast agents (CAs) are applied for better differentiation of a given region of interest (ROI). Currently, many different kinds of inorganic nanoparticulate CAs for spectral CT have been developed due to the limitations of clinical iodine (I)-based contrast media, leading to the conclusion that inorganic nanomedicine applied to spectral CT will be a powerful collaboration both in basic research and in clinics. In this review, the underlying principles and types of spectral CT techniques are discussed, and some evolving clinical diagnosis applications of spectral CT techniques are introduced. In particular, recent developments in inorganic CAs used for spectral CT are summarized. Finally, the challenges and future developments of inorganic nanomedicine in spectral CT are briefly discussed.

(18)F-FDG-PET/CT findings in pancreatic metastasis

PURPOSE

This study aimed to evaluate the fluorodeoxyglucose positron emission tomography/computed tomography (FDG-PET/CT) findings and pattern of FDG uptake in pancreatic metastases.

MATERIALS AND METHODS

A total of 19 consecutive patients (26 lesions) with histologically or clinically confirmed pancreatic metastases who had undergone (18)F-FDG-PET/CT were enrolled in this retrospective study. Among the 19 patients, 14 patients underwent abdominal contrast-enhanced CT (ceCT). The location, size and FDG uptake patterns of the pancreatic lesions were recorded. Metabolic activity by means of maximum standardised uptake value (SUVmax) was measured by drawing regions of interest at the site of pancreatic lesions. Twenty pancreatic cancer patients were included in this study as comparative data analysis. The difference of SUVmax between pancreatic metastases and primary pancreatic cancer were compared using the Mann-Whitney U test. P < 0.05 was considered significant.

RESULTS

Three different patterns of FDG uptake could be distinguished in the pancreatic metastatic lesions, including focal nodule or mass, multiple foci and segmental lesion with high FDG uptake. The average SUVmax in pancreatic metastases was 7.8 ± 6.9 versus 7.4 ± 3.9 in primary pancreatic cancer (P = 0.987 > 0.05). Four intrapancreatic isodense nodules in three patients were undetected on ceCT.

CONCLUSION

The described patterns of FDG uptake findings may be helpful for a better characterisation of pancreatic metastases although semiquantitative analysis using SUVmax could not be used as a criterion for differentiating pancreatic metastases from primary pancreatic cancer. FDG-PET/CT has also an advantage in detecting unsuspected pancreatic metastases which cannot be detected by ceCT imaging. Thus, it is a useful adjunct to the described features on CT.

Cross-modality PET/CT and contrast-enhanced CT imaging for pancreatic cancer

AIM: To explore the diagnostic value of the cross-modality fusion images provided by positron emission tomography/computed tomography (PET/CT) and contrast-enhanced CT (CECT) for pancreatic cancer (PC).

METHODS: Data from 70 patients with pancreatic lesions who underwent CECT and PET/CT examinations at our hospital from August 2010 to October 2012 were analyzed. PET/CECT for the cross-modality image fusion was obtained using TureD software. The diagnostic efficiencies of PET/CT, CECT and PET/CECT were calculated and compared with each other using a χ² test. P < 0.05 was considered to indicate statistical significance.

RESULTS: Of the total 70 patients, 50 had PC and 20 had benign lesions. The differences in the sensitivity, negative predictive value (NPV), and accuracy between CECT and PET/CECT in detecting PC were statistically significant (P < 0.05 for each). In 15 of the 31 patients with PC who underwent a surgical operation, peripancreatic vessel invasion was verified. The differences in the sensitivity, positive predictive value, NPV, and accuracy of CECT vs PET/CT and PET/CECT vs PET/CT in diagnosing peripancreatic vessel invasion were statistically significant (P < 0.05 for each). In 19 of the 31 patients with PC who underwent a surgical operation, regional lymph node metastasis was verified by postsurgical histology. There was no statistically significant difference among the three methods in detecting regional lymph node metastasis (P > 0.05 for each). In 17 of the 50 patients with PC confirmed by histology or clinical follow-up, distant metastasis was confirmed. The differences in the sensitivity and NPV between CECT and PET/CECT in detecting distant metastasis were statistically significant (P < 0.05 for each).

CONCLUSION: Cross-modality image fusion of PET/CT and CECT is a convenient and effective method that can be used to diagnose and stage PC, compensating for the defects of PET/CT and CECT when they are conducted individually.