Improving diagnosis and outcome prediction of gastric cancer via multimodal learning using whole slide pathological images and gene expression.

For the diagnosis and outcome prediction of gastric cancer (GC), machine learning methods based on whole slide pathological images (WSIs) have shown promising performance and reduced the cost of manual analysis. Nevertheless, accurate prediction of GC outcome may rely on multiple modalities with complementary information, particularly gene expression data. Thus, there is a need to develop multimodal learning methods to enhance prediction performance. In this paper, we collect a dataset from Ruijin Hospital and propose a multimodal learning method for GC diagnosis and outcome prediction, called GaCaMML, which is featured by a cross-modal attention mechanism and Per-Slide training scheme. Additionally, we perform feature attribution analysis via integrated gradient (IG) to identify important input features. The proposed method improves prediction accuracy over the single-modal learning method on three tasks, i.e., survival prediction (by 4.9% on C-index), pathological stage classification (by 11.6% on accuracy), and lymph node classification (by 12.0% on accuracy). Especially, the Per-Slide strategy addresses the issue of a high WSI-to-patient ratio and leads to much better results compared with the Per-Person training scheme. For the interpretable analysis, we find that although WSIs dominate the prediction for most samples, there is still a substantial portion of samples whose prediction highly relies on gene expression information. This study demonstrates the great potential of multimodal learning in GC-related prediction tasks and investigates the contribution of WSIs and gene expression, respectively, which not only shows how the model makes a decision but also provides insights into the association between macroscopic pathological phenotypes and microscopic molecular features.

Artificial intelligence-based assessment of PD-L1 expression in diffuse large B cell lymphoma.

Diffuse large B cell lymphoma (DLBCL) is an aggressive blood cancer known for its rapid progression and high incidence. The growing use of immunohistochemistry (IHC) has significantly contributed to the detailed cell characterization, thereby playing a crucial role in guiding treatment strategies for DLBCL. In this study, we developed an AI-based image analysis approach for assessing PD-L1 expression in DLBCL patients. PD-L1 expression represents as a major biomarker for screening patients who can benefit from targeted immunotherapy interventions. In particular, we performed large-scale cell annotations in IHC slides, encompassing over 5101 tissue regions and 146,439 live cells. Extensive experiments in primary and validation cohorts demonstrated the defined quantitative rule helped overcome the difficulty of identifying specific cell types. In assessing data obtained from fine needle biopsies, experiments revealed that there was a higher level of agreement in the quantitative results between Artificial Intelligence (AI) algorithms and pathologists, as well as among pathologists themselves, in comparison to the data obtained from surgical specimens. We highlight that the AI-enabled analytics enhance the objectivity and interpretability of PD-L1 quantification to improve the targeted immunotherapy development in DLBCL patients.

The genetic landscape of histologically transformed marginal zone lymphomas.

BACKGROUND: Marginal zone lymphomas (MZLs) comprise a diverse group of indolent lymphoproliferative disorders; however, some patients develop histologic transformation (HT) with rapid progression to aggressive lymphoma. METHODS: Forty-three MZLs with HT (HT-MZLs), 535 MZLs, and 174 de novo diffuse large B-cell lymphomas (DLBCLs) without rearrangements of MYC, BCL2, and BCL6 were collected. Among these, 22 HT-MZLs, 39 MZLs, and 174 DLBCLs were subjected to 148-gene targeted exome sequencing. The clinicopathologic features of patients who had HT-MZL and their genetic alterations were compared with those of patients who had MZLs and DLBCLs. RESULTS: All 43 HT-MZLs corresponded to DLBCLs. No HT-MZLs harbored BCL2 and MYC and/or BCL6 rearrangements. Bone marrow involvement and higher levels of lactate dehydrogenase were significantly more common in HT-MZLs than in MZLs. Furthermore, upregulated BCL6, MUM1, C-MYC, and Ki-67 expression was observed more frequently in HT-MZLs than in MZLs. TBL1XR1 was the most frequently altered gene (63.6%) in HT-MZLs, followed by CCND3 (31.8%), CARD11, ID3, and TP53 (22.7%). A trend toward worse progression-free survival in patients with TBL1XR1 mutations was observed. Compared with MZLs and non-germinal center B-cell (GCB) type DLBCLs, significantly higher frequencies of TBL1XR1 and ID3 mutations were identified in HT-MZLs. PIM1 mutations frequently occurred in DLBCLs and were significantly associated with TBL1XR1 mutations but were mutated less in HT-MZLs that had TBL1XR1 mutations. CONCLUSIONS: The current findings reveal the clinicopathologic and genetic features of HT-MZLs, suggesting that these tumors might constitute a group distinct from MZL and de novo non-GCB type DLBCL. TBL1XR1 mutations may be considered a predictor of HT in MZL.

Relay learning: a physically secure framework for clinical multi-site deep learning.

Big data serves as the cornerstone for constructing real-world deep learning systems across various domains. In medicine and healthcare, a single clinical site lacks sufficient data, thus necessitating the involvement of multiple sites. Unfortunately, concerns regarding data security and privacy hinder the sharing and reuse of data across sites. Existing approaches to multi-site clinical learning heavily depend on the security of the network firewall and system implementation. To address this issue, we propose Relay Learning, a secure deep-learning framework that physically isolates clinical data from external intruders while still leveraging the benefits of multi-site big data. We demonstrate the efficacy of Relay Learning in three medical tasks of different diseases and anatomical structures, including structure segmentation of retina fundus, mediastinum tumors diagnosis, and brain midline localization. We evaluate Relay Learning by comparing its performance to alternative solutions through multi-site validation and external validation. Incorporating a total of 41,038 medical images from 21 medical hosts, including 7 external hosts, with non-uniform distributions, we observe significant performance improvements with Relay Learning across all three tasks. Specifically, it achieves an average performance increase of 44.4%, 24.2%, and 36.7% for retinal fundus segmentation, mediastinum tumor diagnosis, and brain midline localization, respectively. Remarkably, Relay Learning even outperforms central learning on external test sets. In the meanwhile, Relay Learning keeps data sovereignty locally without cross-site network connections. We anticipate that Relay Learning will revolutionize clinical multi-site collaboration and reshape the landscape of healthcare in the future.

Schwann cells regulate tumor cells and cancer-associated fibroblasts in the pancreatic ductal adenocarcinoma microenvironment.

Neuropathy is a feature more frequently observed in pancreatic ductal adenocarcinoma (PDAC) than other tumors. Schwann cells, the most prevalent cell type in peripheral nerves, migrate toward tumor cells and associate with poor prognosis in PDAC. To unveil the effects of Schwann cells on the neuro-stroma niche, here we perform single-cell RNA-sequencing and microarray-based spatial transcriptome analysis of PDAC tissues. Results suggest that Schwann cells may drive tumor cells and cancer-associated fibroblasts (CAFs) to more malignant subtypes: basal-like and inflammatory CAFs (iCAFs), respectively. Moreover, in vitro and in vivo assays demonstrate that Schwann cells enhance the proliferation and migration of PDAC cells via Midkine signaling and promote the switch of CAFs to iCAFs via interleukin-1α. Culture of tumor cells and CAFs with Schwann cells conditioned medium accelerates PDAC progression. Thus, we reveal that Schwann cells induce malignant subtypes of tumor cells and CAFs in the PDAC milieu.

Altered cytokeratin 5 expression in breast lobular myoepithelial cells.

AIMS: Cytokeratin 5 (CK5) is a surrogate maker of progenitor cells and early glandular and myoepithelial cells (MECs) in the breast, and CK5 expression in breast MECs varies from ducts to lobules, and from normal to diseased tissue. However, the mechanisms underlying immunophenotypic alterations of CK5 expression in MECs remain unclear. METHODS: CK5 expression in MECs of 20 normal breast samples, 58 ductal carcinoma in situ (DCIS; including 21 DCIS with extensive lobular involvement), 11 atypical ductal hyperplasia (ADH), 18 non-invasive lobular neoplasia consisting of 11 atypical lobular hyperplasia (ALH) and 7 lobular carcinoma in situ (LCIS), 20 cystic lobules and 10 usual ductal hyperplasia (UDH) involving lobules were observed to evaluate the effects of contact with benign hyperplastic or cancerous luminal cells and pressure of dilated glands on CK5 expression. RESULTS: CK5 expression in normal ductal MECs was exclusively positive, whereas most normal lobular MECs were negative. In DCIS, cancerous ducts were primarily surrounded by CK5-positive MECs (91.0%), as were lobular acini involved by DCIS (89.2%), while the remaining normal acini maintained CK5-negative. CK5-positive MECs were found in 57.5% of acini in ALH and were more prevalent in LCIS (70.7%). CK5 expression was occasionally positive in both cystic lobules (16.7%) and lobules involved by UDH (14.3%), while an increase of CK5-positive MECs was found in ADH (38.2%). CONCLUSIONS: These results suggest that CK5 expression in lobular MECs may be altered by contact with cancerous luminal cells rather than benign hyperplastic luminal cells or pressure from dilated glands.

Retinol dehydrogenase 10 reduction mediated retinol metabolism disorder promotes diabetic cardiomyopathy in male mice.

Diabetic cardiomyopathy is a primary myocardial injury induced by diabetes with complex pathogenesis. In this study, we identify disordered cardiac retinol metabolism in type 2 diabetic male mice and patients characterized by retinol overload, all-trans retinoic acid deficiency. By supplementing type 2 diabetic male mice with retinol or all-trans retinoic acid, we demonstrate that both cardiac retinol overload and all-trans retinoic acid deficiency promote diabetic cardiomyopathy. Mechanistically, by constructing cardiomyocyte-specific conditional retinol dehydrogenase 10-knockout male mice and overexpressing retinol dehydrogenase 10 in male type 2 diabetic mice via adeno-associated virus, we verify that the reduction in cardiac retinol dehydrogenase 10 is the initiating factor for cardiac retinol metabolism disorder and results in diabetic cardiomyopathy through lipotoxicity and ferroptosis. Therefore, we suggest that the reduction of cardiac retinol dehydrogenase 10 and its mediated disorder of cardiac retinol metabolism is a new mechanism underlying diabetic cardiomyopathy.

Targeting ACSL1 promotes cardiomyocyte proliferation and cardiac regeneration.

BACKGROUND: Neonatal hearts have considerable regenerative potential within 7 days post birth (P7), but the rate of regeneration is extremely low after P7. Interestingly, lipid metabolism increases dramatically after P7. The similarities in these age profiles suggests a possible link between cardiac regeneration and lipid metabolism. Acyl CoA synthase long chain family member 1 (ACSL1) is the key enzyme that regulates lipid metabolism. The aim of this study was to identify the role of ACSL1 in the regeneration of cardiomyocytes. METHODS AND RESULTS: The uptake of fatty acids in hearts increased after P7; however, myocardial regeneration was decreased. We profiled an RNA-sequence array of hearts from mice of different ages, including E10.5 (embryonic stage)-, 3-, 7-, 21-, 30-, and 60-day-old mice, and found that the expression of ACSL1 was significantly increased after P7. By establishing ACSL1 knockdown mice with adeno-associated virus (AAV9). Then, we verified that knockdown of ACSL1 enhanced the capacity for myocardial regeneration both in mice and in primary cardiomyocytes. Indeed, ACSL1 knockdown in primary cardiomyocytes promoted the cell cycle progression from G0 to G2 phase by regulating specific factors, which may correlate with the activation of AKT by ACSL1 and withdrawal of FOXO1 from the nucleus. In vivo, knockdown of ACSL1 effectively restored cardiac function and myocardial regeneration in adult mice with myocardial infarction (MI). CONCLUSIONS: ACSL1 possibly induces the loss of the myocardial regenerative potential beginning at P7 in mice, and inhibition of ACSL1 effectively promoted myocardial repair after MI in mice.

Quantifying the cell morphology and predicting biological behavior of signet ring cell carcinoma using deep learning.

Signet ring cell carcinoma (SRCC) is a malignant tumor of the digestive system. This tumor has long been considered to be poorly differentiated and highly invasive because it has a higher rate of metastasis than well-differentiated adenocarcinoma. But some studies in recent years have shown that the prognosis of some SRCC is more favorable than other poorly differentiated adenocarcinomas, which suggests that SRCC has different degrees of biological behavior. Therefore, we need to find a histological stratification that can predict the biological behavior of SRCC. Some studies indicate that the morphological status of cells can be linked to the invasiveness potential of cells, however, the traditional histopathological examination can not objectively define and evaluate them. Recent improvements in biomedical image analysis using deep learning (DL) based neural networks could be exploited to identify and analyze SRCC. In this study, we used DL to identify each cancer cell of SRCC in whole slide images (WSIs) and quantify their morphological characteristics and atypia. Our results show that the biological behavior of SRCC can be predicted by quantifying the morphology of cancer cells by DL. This technique could be used to predict the biological behavior and may change the stratified treatment of SRCC.

Pigment Epithelial-Derived Factor Deficiency Accelerates Atherosclerosis Development via Promoting Endothelial Fatty Acid Uptake in Mice With Hyperlipidemia.

Background Endothelial cell injury, induced by dyslipidemia, is the initiation of atherosclerosis, resulting in an imbalance in endothelial fatty acid (FA) transport. Pigment epithelial-derived factor (PEDF) is an important regulator in lipid metabolism. We hypothesized that PEDF is involved in endothelium-mediated FA uptake under hyperlipidemic conditions. Methods and Results Circulating PEDF levels were higher in patients with atherosclerotic cardiovascular disease than in normal individuals. However, decreasing trends of serum PEDF levels were confirmed in both wild-type and apolipoprotein E-deficient mice fed a long-term high-fat diet. Apolipoprotein E-deficient/PEDF-deficient mice were generated by crossing PEDF-deficient mice with apolipoprotein E-deficient mice, and then mice were fed with 24, 36, or 48 weeks of high-fat diet. Greater increases in body fat and plasma lipids were displayed in PEDF-deficient mice. In addition, PEDF deficiency in mice accelerated atherosclerosis, as evidenced by increased atherosclerotic plaques, pronounced vascular dysfunction, and increased lipid accumulation in peripheral tissues, whereas injection of adeno-associated virus encoding PEDF exerted opposite effects. Mechanistically, PEDF inhibited the vascular endothelial growth factor B paracrine signaling by reducing secretion of protein vascular endothelial growth factor B in peripheral tissue cells and decreasing expression of its downstream targets in endothelial cells, including its receptors (namely, vascular endothelial growth factor receptor-1 and neuropilin-1), and FA transport proteins 3 and 4, to suppress endothelial FA uptake, whereas PEDF deletion in mice activated the vascular endothelial growth factor B signaling pathway, thus causing markedly increased lipid accumulation. Conclusions Decreasing expression of PEDF aggravates atherosclerosis by significantly impaired vascular function and enhanced endothelial FA uptake, thus exacerbating ectopic lipid deposition in peripheral tissues.