The deep learning prediction model has the potential to predict lymph node metastasis in patients with clinically lymph node–negative breast cancer on the basis of US images of primary breast cancer. Background Deep learning (DL) algorithms are gaining extensive attention for their excellent performance in image recognition tasks. DL models can automatically make a quantitative assessment of co...

Objective . The incidence of primary thyroid cancer has risen steadily over the past decades because of overdiagnosis and overtreatment through the improvement in imaging techniques for screening, especially in ultrasound examination. Metastatic status of lymph nodes is important for staging the type of primary thyroid cancer. Deep learning algorithms based on ultrasound images were thus developed to assist radiologists on the diagnosis of lymph node metastasis. The objective of this study is to integrate more clinical context (e.g., health records and various image modalities) into, and explore more interpretable patterns discovered by, deep learning algorithms for the prediction of lymph node metastasis in primary thyroid cancer patients. Approach . A deep multimodal learning network was developed in this study with a novel index proposed to compare the contribution of different modalities when making the predictions. Main results . The proposed multimodal network achieved an average F1 score of 0.888 and an average area under the receiver operating characteristic curve (AUC) value of 0.973 in two independent validation sets, and the performance was significantly better than that of three single-modality deep learning networks. Moreover, among three modalities used in this study, the deep multimodal learning network relied generally more on image modalities than the data modality of clinic records when making the predictions. Significance . Our work is beneficial to prospective clinic trials of radiologists on the diagnosis of lymph node metastasis in primary thyroid cancer, and will better help them understand how the predictions are made in deep multimodal learning algorithms.

Background: Non-invasive risk stratification contributes to the precise treatment of prostate cancer (PCa). In previous studies, lymphocyte subsets were used to differentiate between low-/intermediate-risk and high-risk PCa, with limited clinical value and poor interpretability. Based on functional subsets of peripheral lymphocyte with the largest sample size to date, this study aims to construct an easy-to-use and robust nomogram to guide the tripartite risk stratifications for PCa. Methods: We retrospectively collected data from 2039 PCa and benign prostate disease (BPD) patients with 42 clinical characteristics on functional subsets of peripheral lymphocyte. After quality control and feature selection, clinical data with the optimal feature subset were utilized for the 10-fold cross-validation of five Machine Learning (ML) models for the task of predicting low-, intermediate- and high-risk stratification of PCa. Then, a novel clinic-ML nomogram was constructed using probabilistic predictions of the trained ML models via the combination of a multivariable Ordinal Logistic Regression analysis and the proposed feature mapping algorithm. Results: 197 PCa patients, including 56 BPD, were enrolled in the study. An optimal subset with nine clinical features was selected. Compared with the best ML model and the clinic nomogram, the clinic-ML nomogram achieved the superior performance with a sensitivity of 0.713 (95% CI 0.573–0.853), specificity of 0.869 (95% CI 0.764–0.974), F1 of 0.699 (95% CI 0.557–0.841), and AUC of 0.864 (95% CI 0.794–0.935). The calibration curve and Decision Curve Analysis (DCA) indicated the predictive capacity and net benefits of the clinic-ML nomogram were improved. Conclusion: Combining the interpretability and simplicity of a nomogram with the efficacy and robustness of ML models, the proposed clinic-ML nomogram can serve as an insight tool for preoperative assessment of PCa risk stratifications, and could provide essential information for the individual diagnosis and treatment in PCa patients.