Accurate preoperative staging and HER2 status prediction of gastric cancer by the deep learning system based on enhanced computed tomography

CT-based multi-regional radiomics model for predicting contrast medium extravasation in patients with tumors: A case-control study

OBJECTIVE

To develop a non-contrast CT based multi-regional radiomics model for predicting contrast medium (CM) extravasation in patients with tumors.

METHODS

A retrospective analysis of non-contrast CT scans from 282 tumor patients across two medical centers led to the development of a radiomics model, using 157 patients for training, 68 for validation, and 57 from an external center as an independent test cohort. The different volumes of interest from right common carotid artery/right internal jugular vein, right subclavian artery/vein and thoracic aorta were delineated. Radiomics features from the training cohort were used to calculate radiomics scores (Rad scores) and develop radiomics model. Non-contrast CT radiomics features were combined with clinical factors to develop an integrated model. A nomogram was created to visually represent the integration of radiomic signatures and clinical factors. The model's predictive performance and clinical utility were evaluated using receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA), respectively. Calibration curves were also used to assess the concordance between the model-predicted probabilities and the observed event probabilities.

RESULTS

Thirteen radiomics features were selected to determine the Rad score. The radiomic model outperformed the clinical model in the training, validation, and external test cohorts, achieving a greater area under the ROC curve (AUC) with values of 0.877, 0.866, 0.828 compared to the clinical model's 0.852, 0.806, 0.740. The combined model yielded better AUC of 0.945, 0.911, and 0.869 in the respective cohorts. The nomogram identified females, the elderly, individuals with hypertension, long term chemotherapy, radiomic signatures as independent risk factors for CM extravasation in patients with tumors. Calibration and DCA validated the high accuracy and clinical utility of this model.

CONCLUSIONS

Radiomics models based on multi-regional non-contrast CT image offered improved prediction of CM extravasation compared with clinical model alone.

Deep learning-based computed tomography urography image analysis for prediction of HER2 status in bladder cancer

Purpose: Bladder cancer (BCa) is one of the most common malignant tumors in the urinary system. BCa with HER2 overexpression can benefit from more precise treatments, but HER2 testing is costly and subjective. This study aimed to detect HER2 expression using computed tomography urography (CTU) images. Method: We gathered CTU images from 97 patients with BCa during the excretion phase in Renmin Hospital of Wuhan University, manually delineated the BCa regions, extracted radiomic features using the Pyradiomics package, conducted data dimensionality reduction via principal component analysis (PCA), and trained three models (Logistic Regression [LR], Random Forest [RF] and Multilayer Perceptron [MLP]) to discern the HER2 expression status. Results: Pyradiomics package was used to extract 975 radiological features and the cumulative interpretation area under the variance curve was 90.964 by PCA. Using an MLP-based deep learning model for identifying HER2 overexpression, the area under the curve (AUC) reached 0.79 (95% confidence interval [CI] 0.74-0.86) in the training set and 0.73 (95% CI 0.66-0.77) in the validation set. LR and RF had AUC of 0.69 (95% CI 0.63-0.75) and 0.66 (95% CI 0.61-0.70) in the training set and 0.61 (95% CI 0.55-0.67) and 0.59 (95% CI 0.55-0.63) in the test set, respectively. Conclusion: The study firstly presents a non-invasive method for identifying and detecting HER2 expression in BCa CTU images. It might not only reduce the cost and subjectivity of traditional HER2 testing but also provide a new technical method for the precise treatment of BCa.

Development of a deep learning model for early gastric cancer diagnosis using preoperative computed tomography images

Background

Gastric cancer is a highly prevalent and fatal disease. Accurate differentiation between early gastric cancer (EGC) and advanced gastric cancer (AGC) is essential for personalized treatment. Currently, the diagnostic accuracy of computerized tomography (CT) for gastric cancer staging is insufficient to meet clinical requirements. Many studies rely on manual marking of lesion areas, which is not suitable for clinical diagnosis.

Methods

In this study, we retrospectively collected data from 341 patients with gastric cancer at the First Affiliated Hospital of Wenzhou Medical University. The dataset was randomly divided into a training set (n=273) and a validation set (n=68) using an 8:2 ratio. We developed a two-stage deep learning model that enables fully automated EGC screening based on CT images. In the first stage, an unsupervised domain adaptive segmentation model was employed to automatically segment the stomach on unlabeled portal phase CT images. Subsequently, based on the results of the stomach segmentation model, the image was cropped out of the stomach area and scaled to a uniform size, and then the EGC and AGC classification models were built based on these images. The segmentation accuracy of the model was evaluated using the dice index, while the classification performance was assessed using metrics such as the area under the curve (AUC) of the receiver operating characteristic (ROC), accuracy, sensitivity, specificity, and F1 score.

Results

The segmentation model achieved an average dice accuracy of 0.94 on the hand-segmented validation set. On the training set, the EGC screening model demonstrated an AUC, accuracy, sensitivity, specificity, and F1 score of 0.98, 0.93, 0.92, 0.92, and 0.93, respectively. On the validation set, these metrics were 0.96, 0.92, 0.90, 0.89, and 0.93, respectively. After three rounds of data regrouping, the model consistently achieved an AUC above 0.9 on both the validation set and the validation set.

Conclusion

The results of this study demonstrate that the proposed method can effectively screen for EGC in portal venous CT images. Furthermore, the model exhibits stability and holds promise for future clinical applications.