Detecting microsatellite instability (MSI) in colorectal cancer is crucial for clinical decision making, as it identifies patients with differential treatment response and prognosis. Universal MSI ...testing is recommended, but many patients remain untested. A critical need exists for broadly accessible, cost-efficient tools to aid patient selection for testing. Here, we investigate the potential of a deep learning-based system for automated MSI prediction directly from haematoxylin and eosin (H&E)-stained whole-slide images (WSIs).
Our deep learning model (MSINet) was developed using 100 H&E-stained WSIs (50 with microsatellite stability MSS and 50 with MSI) scanned at 40× magnification, each from a patient randomly selected in a class-balanced manner from the pool of 343 patients who underwent primary colorectal cancer resection at Stanford University Medical Center (Stanford, CA, USA; internal dataset) between Jan 1, 2015, and Dec 31, 2017. We internally validated the model on a holdout test set (15 H&E-stained WSIs from 15 patients; seven cases with MSS and eight with MSI) and externally validated the model on 484 H&E-stained WSIs (402 cases with MSS and 77 with MSI; 479 patients) from The Cancer Genome Atlas, containing WSIs scanned at 40× and 20× magnification. Performance was primarily evaluated using the sensitivity, specificity, negative predictive value (NPV), and area under the receiver operating characteristic curve (AUROC). We compared the model's performance with that of five gastrointestinal pathologists on a class-balanced, randomly selected subset of 40× magnification WSIs from the external dataset (20 with MSS and 20 with MSI).
The MSINet model achieved an AUROC of 0·931 (95% CI 0·771–1·000) on the holdout test set from the internal dataset and 0·779 (0·720–0·838) on the external dataset. On the external dataset, using a sensitivity-weighted operating point, the model achieved an NPV of 93·7% (95% CI 90·3–96·2), sensitivity of 76·0% (64·8–85·1), and specificity of 66·6% (61·8–71·2). On the reader experiment (40 cases), the model achieved an AUROC of 0·865 (95% CI 0·735–0·995). The mean AUROC performance of the five pathologists was 0·605 (95% CI 0·453–0·757).
Our deep learning model exceeded the performance of experienced gastrointestinal pathologists at predicting MSI on H&E-stained WSIs. Within the current universal MSI testing paradigm, such a model might contribute value as an automated screening tool to triage patients for confirmatory testing, potentially reducing the number of tested patients, thereby resulting in substantial test-related labour and cost savings.
Stanford Cancer Institute and Stanford Departments of Pathology and Biomedical Data Science.
BRAF mutation in colorectal cancer is associated with microsatellite instability (MSI) through its relationship with high-level CpG island methylator phenotype (CIMP) and MLH1 promoter methylation. ...MSI and BRAF mutation analyses are routinely used for familial cancer risk assessment. To clarify clinical outcome associations of combined MSI/BRAF subgroups, we investigated survival in 1253 rectal and colon cancer patients within the Nurses' Health Study and Health Professionals Follow-up Study with available data on clinical and other molecular features, including CIMP, LINE-1 hypomethylation, and KRAS and PIK3CA mutations. Compared with the majority subtype of microsatellite stable (MSS)/BRAF-wild-type, MSS/BRAF-mutant, MSI-high/BRAF-mutant, and MSI-high/BRAF-wild-type subtypes showed multivariable colorectal cancer-specific mortality hazard ratios of 1.60 (95% confidence interval CI =1.12 to 2.28; P = .009), 0.48 (95% CI = 0.27 to 0.87; P = .02), and 0.25 (95% CI = 0.12 to 0.52; P < .001), respectively. No evidence existed for a differential prognostic role of BRAF mutation by MSI status (P(interaction) > .50). Combined BRAF/MSI status in colorectal cancer is a tumor molecular biomarker for prognosic risk stratification.
Objective Postoperative atrial fibrillation is the most common complication after cardiac surgery. A variety of postoperative atrial fibrillation risk factors have been reported, but study results ...have been inconsistent or contradictory, particularly in patients with preexisting atrial fibrillation. The incidence of postoperative atrial fibrillation was evaluated in a group of 10,390 patients undergoing cardiac surgery among a comprehensive range of risk factors to identify reliable predictors of postoperative atrial fibrillation. Methods This 20-year retrospective study examined the relationship between postoperative atrial fibrillation and demographic factors, preoperative health conditions and medications, operative procedures, and postoperative complications. Multivariate logistic regression models were used to evaluate potential predictors of postoperative atrial fibrillation. Results Increasing age, mitral valve surgery (odds ratio = 1.91), left ventricular aneurysm repair (odds ratio = 1.57), aortic valve surgery (odds ratio = 1.52), race (Caucasian) (odds ratio = 1.51), use of cardioplegia (odds ratio = 1.36), use of an intraaortic balloon pump (odds ratio = 1.28), previous congestive heart failure (odds ratio = 1.28), and hypertension (odds ratio = 1.15) were significantly associated with postoperative atrial fibrillation. The non-linear relationship between age and postoperative atrial fibrillation revealed the acceleration of postoperative atrial fibrillation risk in patients aged 55 years or more. In patients undergoing coronary artery bypass grafting, increasing age and previous congestive heart failure were the only factors associated with a higher risk of postoperative atrial fibrillation. There was no trend in incidence of postoperative atrial fibrillation over time. No protective factors against postoperative atrial fibrillation were detected, including commonly prescribed categories of medications. Conclusions The persistence of the problem of postoperative atrial fibrillation and the modest predictability using common risk factors suggest that limited progress has been made in understanding its cause and treatment.
Recurrence risk stratification of patients undergoing primary surgical resection for hepatocellular carcinoma (HCC) is an area of active investigation, and several staging systems have been proposed ...to optimize treatment strategies. However, as many as 70% of patients still experience tumor recurrence at 5 years post-surgery. We developed and validated a deep learning-based system (HCC-SurvNet) that provides risk scores for disease recurrence after primary resection, directly from hematoxylin and eosin-stained digital whole-slide images of formalin-fixed, paraffin embedded liver resections. Our model achieved concordance indices of 0.724 and 0.683 on the internal and external test cohorts, respectively, exceeding the performance of the standard Tumor-Node-Metastasis classification system. The model's risk score stratified patients into low- and high-risk subgroups with statistically significant differences in their survival distributions, and was an independent risk factor for post-surgical recurrence in both test cohorts. Our results suggest that deep learning-based models can provide recurrence risk scores which may augment current patient stratification methods and help refine the clinical management of patients undergoing primary surgical resection for HCC.
The temporal switch from progenitor cell proliferation to differentiation is essential for effective adult tissue repair. We previously reported the critical role of Notch signaling in the ...proliferative expansion of myogenic progenitors in mammalian postnatal myogenesis. We now show that the onset of differentiation is due to a transition from Notch signaling to Wnt signaling in myogenic progenitors and is associated with an increased expression of Wnt in the tissue and an increased responsiveness of progenitors to Wnt. Crosstalk between these two pathways occurs via GSK3beta, which is maintained in an active form by Notch but is inhibited by Wnt in the canonical Wnt signaling cascade. These results demonstrate that the temporal balance between Notch and Wnt signaling orchestrates the precise progression of muscle precursor cells along the myogenic lineage pathway, through stages of proliferative expansion and then differentiation, during postnatal myogenesis.
is commonly mutated in pancreatic ductal adenocarcinoma (PDAC), but the functional effects of
mutations in the pancreas are unclear. Understanding the molecular mechanisms that drive PDAC formation ...may lead to novel therapies.
Concurrent conditional
deletion and
activation mutations were modelled in mice. Small-interfering RNA (siRNA) and CRISPR/Cas9 were used to abrogate
in human pancreatic ductal epithelial cells.
We found that pancreas-specific
loss in mice was sufficient to induce inflammation, pancreatic intraepithelial neoplasia (PanIN) and mucinous cysts. Concurrent
activation accelerated the development of cysts that resembled intraductal papillary mucinous neoplasm. Lineage-specific
deletion confirmed compartment-specific tumour-suppressive effects. Duct-specific
loss promoted dilated ducts with occasional cyst and PDAC formation. Heterozygous acinar-specific
loss resulted in accelerated PanIN and PDAC formation with worse survival. RNA-seq showed that
loss induced gene networks associated with
activity and protein translation.
knockdown in human pancreatic ductal epithelial cells induced increased MYC expression and protein synthesis that was abrogated with
knockdown. ChIP-seq against H3K27ac demonstrated an increase in activated enhancers/promoters.
suppresses pancreatic neoplasia in a compartment-specific manner. In duct cells, this process appears to be associated with MYC-facilitated protein synthesis.
Early identification of gastric precancerous lesions, including atrophic gastritis (AG) and intestinal metaplasia (IM), may improve gastric cancer detection and prevention. Because AG and IM are ...generally asymptomatic, many of the estimated 15 million Americans who carry these lesions remain undiagnosed.
AG and IM are associated with either active or prior Helicobacter pylori (Hp) infection. Hp infection leads to perturbations in the serum concentration of gastric hormones pepsinogen I (PGI), pepsinogen II, the pepsinogen I/II ratio (PGR), gastrin-17 (G-17), and Hp IgG.
In East Asia and other regions with high burden of Hp infection and gastric cancer, these biomarkers have been used as screening tools for AG and IM.
However, there exists limited data on the sensitivity and discrimination of these serologic markers in low-Hp-prevalence populations, such as the United States.
Artificial intelligence (AI) algorithms continue to rival human performance on a variety of clinical tasks, while their actual impact on human diagnosticians, when incorporated into clinical ...workflows, remains relatively unexplored. In this study, we developed a deep learning-based assistant to help pathologists differentiate between two subtypes of primary liver cancer, hepatocellular carcinoma and cholangiocarcinoma, on hematoxylin and eosin-stained whole-slide images (WSI), and evaluated its effect on the diagnostic performance of 11 pathologists with varying levels of expertise. Our model achieved accuracies of 0.885 on a validation set of 26 WSI, and 0.842 on an independent test set of 80 WSI. Although use of the assistant did not change the mean accuracy of the 11 pathologists (
= 0.184, OR = 1.281), it significantly improved the accuracy (
= 0.045, OR = 1.499) of a subset of nine pathologists who fell within well-defined experience levels (GI subspecialists, non-GI subspecialists, and trainees). In the assisted state, model accuracy significantly impacted the diagnostic decisions of all 11 pathologists. As expected, when the model's prediction was correct, assistance significantly improved accuracy (
= 0.000, OR = 4.289), whereas when the model's prediction was incorrect, assistance significantly decreased accuracy (
= 0.000, OR = 0.253), with both effects holding across all pathologist experience levels and case difficulty levels. Our results highlight the challenges of translating AI models into the clinical setting, and emphasize the importance of taking into account potential unintended negative consequences of model assistance when designing and testing medical AI-assistance tools.
Previous approaches to defining subtypes of colorectal carcinoma (CRC) and other cancers based on transcriptomes have assumed the existence of discrete subtypes. We analyze gene expression patterns ...of colorectal tumors from a large number of patients to test this assumption and propose an approach to identify potentially a continuum of subtypes that are present across independent studies and cohorts.
We examine the assumption of discrete CRC subtypes by integrating 18 published gene expression datasets and > 3700 patients, and contrary to previous reports, find no evidence to support the existence of discrete transcriptional subtypes. Using a meta-analysis approach to identify co-expression patterns present in multiple datasets, we identify and define robust, continuously varying subtype scores to represent CRC transcriptomes. The subtype scores are consistent with established subtypes (including microsatellite instability and previously proposed discrete transcriptome subtypes), but better represent overall transcriptional activity than do discrete subtypes. The scores are also better predictors of tumor location, stage, grade, and times of disease-free survival than discrete subtypes. Gene set enrichment analysis reveals that the subtype scores characterize T-cell function, inflammation response, and cyclin-dependent kinase regulation of DNA replication.
We find no evidence to support discrete subtypes of the CRC transcriptome and instead propose two validated scores to better characterize a continuity of CRC transcriptomes.
Enteric glia are a distinct population of peripheral glial cells in the enteric nervous system that regulate intestinal homeostasis, epithelial barrier integrity, and gut defense. Given these unique ...attributes, we investigated the impact of enteric glia depletion on tumor development in azoxymethane/dextran sodium sulfate (AOM/DSS)-treated mice, a classical model of colorectal cancer (CRC). Depleting GFAP
enteric glia resulted in a profoundly reduced tumor burden in AOM/DSS mice and additionally reduced adenomas in the
mouse model of familial adenomatous polyposis, suggesting a tumor-promoting role for these cells at an early premalignant stage. This was confirmed in further studies of AOM/DSS mice, as enteric glia depletion did not affect the properties of established malignant tumors but did result in a marked reduction in the development of precancerous dysplastic lesions. Surprisingly, the protective effect of enteric glia depletion was not dependent on modulation of anti-tumor immunity or intestinal inflammation. These findings reveal that GFAP
enteric glia play a critical pro-tumorigenic role during early CRC development and identify these cells as a potential target for CRC prevention.