Predicting 5-year recurrence risk in colorectal cancer: development and validation of a histology-based deep learning approach.

BACKGROUND: Accurate estimation of the long-term risk of recurrence in patients with non-metastatic colorectal cancer (CRC) is crucial for clinical management. Histology-based deep learning is expected to provide more abundant information for risk stratification. METHODS: We developed and validated a weakly supervised deep-learning model for predicting 5-year relapse-free survival (RFS) to stratify patients with different risks based on histological images from three hospitals of 614 cases with non-metastatic CRC. A deep prognostic factor (DL-RRS) was established to stratify patients into high and low-risk group. The areas under the curve (AUCs) were calculated to evaluate the performances of models. RESULTS: Our proposed model achieves the AUCs of 0.833 (95% CI: 0.736-0.905) and 0.715 (95% CI: 0.647-0.776) on validation cohort and external test cohort, respectively. The 5-year RFS rate was 45.7% for high DL-RRS patients, and 82.5% for low DL-RRS patients respectively in the external test cohort (HR: 3.89, 95% CI: 2.51-6.03, P < 0.001). Adjuvant chemotherapy was associated with improved RFS in Stage II patients with high DL-RRS (HR: 0.15, 95% CI: 0.06-0.38, P < 0.001). CONCLUSIONS: DL-RRS has a good predictive performance of 5-year recurrence risk in CRC, and will better serve the clinical decision-making.

Identification of DNA methylation signatures for hepatocellular carcinoma detection and microvascular invasion prediction.

BACKGROUND AND AIM: Preoperative evaluation of microvascular invasion (MVI) in patients with hepatocellular carcinoma (HCC) is important for surgical strategy determination. We aimed to develop and establish a preoperative predictive model for MVI status based on DNA methylation markers. METHODS: A total of 35 HCC tissues and the matched peritumoral normal liver tissues as well as 35 corresponding HCC patients' plasma samples and 24 healthy plasma samples were used for genome-wide methylation sequencing and subsequent methylation haplotype block (MHB) analysis. Predictive models were constructed based on selected MHB markers and 3-cross validation was used. RESULTS: We grouped 35 HCC patients into 2 categories, including the MVI- group with 17 tissue and plasma samples, and MVI + group with 18 tissue and plasma samples. We identified a tissue DNA methylation signature with an AUC of 98.0% and a circulating free DNA (cfDNA) methylation signature with an AUC of 96.0% for HCC detection. Furthermore, we established a tissue DNA methylation signature for MVI status prediction, and achieved an AUC of 85.9%. Based on the MVI status predicted by the DNA methylation signature, the recurrence-free survival (RFS) and overall survival (OS) were significantly better in the predicted MVI- group than that in the predicted MVI + group. CONCLUSIONS: In this study, we identified a cfDNA methylation signature for HCC detection and a tissue DNA methylation signature for MVI status prediction with high accuracy.

Deep learning for evaluation of microvascular invasion in hepatocellular carcinoma from tumor areas of histology images.

BACKGROUND: Microvascular invasion (MVI) is essential for the management of hepatocellular carcinoma (HCC). However, MVI is hard to evaluate in patients without sufficient peri-tumoral tissue samples, which account for over a half of HCC patients. METHODS: We established an MVI deep-learning (MVI-DL) model with a weakly supervised multiple-instance learning framework, to evaluate MVI status using only tumor tissues from the histological whole slide images (WSIs). A total of 350 HCC patients (2917 WSIs) from the First Affiliated Hospital of Sun Yat-sen University (FAHSYSU cohort) were divided into a training and test set. One hundred and twenty patients (504 WSIs) from Dongguan People's Hospital and Shunde Hospital of Southern Medical University (DG-SD cohort) formed an external test set. Unsupervised clustering and class activation mapping were applied to visualize the key histological features. RESULTS: In the FAHSYSU and DG-SD test set, the MVI-DL model achieved an AUC of 0.904 (95% CI 0.888-0.920) and 0.871 (95% CI 0.837-0.905), respectively. Visualization results showed that macrotrabecular architecture with rich blood sinus, rich tumor stroma and high intratumor heterogeneity were identified as the key features associated with MVI ( +), whereas severe immune infiltration and highly differentiated tumor cells were associated with MVI (-). In the simulation of patients with only one WSI or biopsies only, the AUC of the MVI-DL model reached 0.875 (95% CI 0.855-0.895) and 0.879 (95% CI 0.853-0.906), respectively. CONCLUSION: The effective, interpretable MVI-DL model has potential as an important tool with practical clinical applicability in evaluating MVI status from the tumor areas on the histological slides.

Integrative metabolomic characterization identifies plasma metabolomic signature in the diagnosis of papillary thyroid cancer.

Discrimination of malignancy from thyroid nodules poses challenges in clinical practice. We aimed to identify the plasma metabolomic biomarkers in discriminating papillary thyroid cancer (PTC) from benign thyroid nodule (BTN). Metabolomics profiling of plasma was performed in two independent cohorts of 651 subjects of PTC (n = 215), BTN (n = 230), and healthy controls (n = 206). In addition, 132 patients with thyroid micronodules (<1 cm) and 44 patients with BTN suspected malignancy by ultrasound were used for biomarker validation. Recursive feature elimination algorithm was used for metabolic biomarkers selecting. Significant differential metabolites were demonstrated in patients with thyroid nodules (PTC and BTN) from healthy controls (P = 0.0001). A metabolic biomarker panel (17 differential metabolites) was identified to discriminate PTC from BTN with an AUC of 97.03% (95% CI: 95.28-98.79%), 91.89% sensitivity, and 92.63% specificity in discovery cohort. The panel had an AUC of 92.72% (95% CI: 87.46-97.99%), 86.57% sensitivity, and 92.50% specificity in validation cohort. The metabolic biomarker signature could correctly identify 84.09% patients whose nodules were suspected malignant by ultrasonography but finally histological benign. Moreover, high accuracy of 87.88% for diagnosis of papillary thyroid microcarcinoma was displayed by this panel and showed significant improvement in accuracy, AUC and specificity when compared with ultrasound. We identified a novel metabolic biomarker signature to discriminate PTC from BTN. The clinical use of this biomarker panel would have improved diagnosis stratification of thyroid microcarcinoma in comparison to ultrasound.