Development and Validation of a Multimodality Model Based on Whole-Slide Imaging and Biparametric MRI for Predicting Postoperative Biochemical Recurrence in Prostate Cancer.

Purpose To develop and validate a machine learning multimodality model based on preoperative MRI, surgical whole-slide imaging (WSI), and clinical variables for predicting prostate cancer (PCa) biochemical recurrence (BCR) following radical prostatectomy (RP). Materials and Methods In this retrospective study (September 2015 to April 2021), 363 male patients with PCa who underwent RP were divided into training (n = 254; median age, 69 years [IQR, 64-74 years]) and testing (n = 109; median age, 70 years [IQR, 65-75 years]) sets at a ratio of 7:3. The primary end point was biochemical recurrence-free survival. The least absolute shrinkage and selection operator Cox algorithm was applied to select independent clinical variables and construct the clinical signature. The radiomics signature and pathomics signature were constructed using preoperative MRI and surgical WSI data, respectively. A multimodality model was constructed by combining the radiomics signature, pathomics signature, and clinical signature. Using Harrell concordance index (C index), the predictive performance of the multimodality model for BCR was assessed and compared with all single-modality models, including the radiomics signature, pathomics signature, and clinical signature. Results Both radiomics and pathomics signatures achieved good performance for BCR prediction (C index: 0.742 and 0.730, respectively) on the testing cohort. The multimodality model exhibited the best predictive performance, with a C index of 0.860 on the testing set, which was significantly higher than all single-modality models (all P ≤ .01). Conclusion The multimodality model effectively predicted BCR following RP in patients with PCa and may therefore provide an emerging and accurate tool to assist postoperative individualized treatment. Keywords: MR Imaging, Urinary, Pelvis, Comparative Studies Supplemental material is available for this article. © RSNA, 2024.

The practical clinical role of machine learning models with different algorithms in predicting prostate cancer local recurrence after radical prostatectomy.

BACKGROUND: The detection of local recurrence for prostate cancer (PCa) patients following radical prostatectomy (RP) is challenging and can influence the treatment plan. Our aim was to construct and verify machine learning models with three different algorithms based on post-operative mpMRI for predicting local recurrence of PCa after RP and explore their potential clinical value compared with the Prostate Imaging for Recurrence Reporting (PI-RR) score of expert-level radiologists. METHODS: A total of 176 patients were retrospectively enrolled and randomly divided into training (n = 123) and testing (n = 53) sets. The PI-RR assessments were performed by two expert-level radiologists with access to the operative histopathological and pre-surgical clinical results. The radiomics models to predict local recurrence were built by utilizing three different algorithms (i.e., support vector machine [SVM], linear discriminant analysis [LDA], and logistic regression-least absolute shrinkage and selection operator [LR-LASSO]). The combined model integrating radiomics features and PI-RR score was developed using the most effective classifier. The classification performances of the proposed models were assessed by receiver operating characteristic (ROC) curve analysis. RESULTS: There were no significant differences between the training and testing sets concerning age, prostate-specific antigen (PSA), Gleason score, T-stage, seminal vesicle invasion (SVI), perineural invasion (PNI), and positive surgical margins (PSM). The radiomics model based on LR-LASSO exhibited superior performance than other radiomics models, with an AUC of 0.858 in the testing set; the PI-RR yielded an AUC of 0.833, and there was no significant difference between the best radiomics model and the PI-RR score. The combined model achieved the best predictive performance with an AUC of 0.924, and a significant difference was observed between the combined model and PI-RR score. CONCLUSIONS: Our radiomics model is an effective tool to predict PCa local recurrence after RP. By integrating radiomics features with the PI-RR score, our combined model exhibited significantly better predictive performance of local recurrence than expert-level radiologists' PI-RR assessment.

Automatic Modulation Classification Using Hybrid Data Augmentation and Lightweight Neural Network.

Automatic modulation classification (AMC) plays an important role in intelligent wireless communications. With the rapid development of deep learning in recent years, neural network-based automatic modulation classification methods have become increasingly mature. However, the high complexity and large number of parameters of neural networks make them difficult to deploy in scenarios and receiver devices with strict requirements for low latency and storage. Therefore, this paper proposes a lightweight neural network-based AMC framework. To improve classification performance, the framework combines complex convolution with residual networks. To achieve a lightweight design, depthwise separable convolution is used. To compensate for any performance loss resulting from a lightweight design, a hybrid data augmentation scheme is proposed. The simulation results demonstrate that the lightweight AMC framework reduces the number of parameters by approximately 83.34% and the FLOPs by approximately 83.77%, without a degradation in performance.

Utility of Clinical-Radiomic Model to Identify Clinically Significant Prostate Cancer in Biparametric MRI PI-RADS V2.1 Category 3 Lesions.

Purpose: To determine the predictive performance of the integrated model based on clinical factors and radiomic features for the accurate identification of clinically significant prostate cancer (csPCa) among Prostate Imaging Reporting and Data System (PI-RADS) 3 lesions. Materials and Methods: A retrospective study of 103 patients with PI-RADS 3 lesions who underwent pre-operative 3.0-T MRI was performed. Patients were randomly divided into the training set and the testing set at a ratio of 7:3. Radiomic features were extracted from axial T2WI, diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC) images of each patient. The minimum redundancy maximum relevance (mRMR) and least absolute shrinkage and selection operator (LASSO) feature selection methods were used to identify the radiomic features and construct a radiomic model for csPCa identification. Moreover, multivariable logistic regression analysis was used to integrate the clinical factors with radiomic feature model to further improve the accuracy of csPCa identification, and the two are presented in the form of normogram. The performance of the integrated model was compared with radiomic model and clinical model on testing set. Results: A total of four radiomic features were selected and used for radiomic model construction producing a radiomic score (Radscore). Radscore was significantly different between the csPCa and the non-csPCa patients (training set: p < 0.001; testing set: p = 0.035). Multivariable logistic regression analysis showed that age and PSA could be used as independent predictors for csPCa identification. The clinical-radiomic model produced the receiver operating characteristic (ROC) curve (AUC) in the testing set was 0.88 (95%CI, 0.75-1.00), which was similar to clinical model (AUC = 0.85; 95%CI, 0.52-0.90) (p = 0.048) and higher than the radiomic model (AUC = 0.71; 95%CI, 0.68-1.00) (p < 0.001). The decision curve analysis implies that the clinical-radiomic model could be beneficial in identifying csPCa among PI-RADS 3 lesions. Conclusion: The clinical-radiomic model could effectively identify csPCa among biparametric PI-RADS 3 lesions and thus could help avoid unnecessary biopsy and improve the life quality of patients.