Deep learning-based risk stratification of preoperative breast biopsies using digital whole slide images.

BACKGROUND: Nottingham histological grade (NHG) is a well established prognostic factor in breast cancer histopathology but has a high inter-assessor variability with many tumours being classified as intermediate grade, NHG2. Here, we evaluate if DeepGrade, a previously developed model for risk stratification of resected tumour specimens, could be applied to risk-stratify tumour biopsy specimens. METHODS: A total of 11,955,755 tiles from 1169 whole slide images of preoperative biopsies from 896 patients diagnosed with breast cancer in Stockholm, Sweden, were included. DeepGrade, a deep convolutional neural network model, was applied for the prediction of low- and high-risk tumours. It was evaluated against clinically assigned grades NHG1 and NHG3 on the biopsy specimen but also against the grades assigned to the corresponding resection specimen using area under the operating curve (AUC). The prognostic value of the DeepGrade model in the biopsy setting was evaluated using time-to-event analysis. RESULTS: Based on preoperative biopsy images, the DeepGrade model predicted resected tumour cases of clinical grades NHG1 and NHG3 with an AUC of 0.908 (95% CI: 0.88; 0.93). Furthermore, out of the 432 resected clinically-assigned NHG2 tumours, 281 (65%) were classified as DeepGrade-low and 151 (35%) as DeepGrade-high. Using a multivariable Cox proportional hazards model the hazard ratio between DeepGrade low- and high-risk groups was estimated as 2.01 (95% CI: 1.06; 3.79). CONCLUSIONS: DeepGrade provided prediction of tumour grades NHG1 and NHG3 on the resection specimen using only the biopsy specimen. The results demonstrate that the DeepGrade model can provide decision support to identify high-risk tumours based on preoperative biopsies, thus improving early treatment decisions.

Clinical evaluation of deep learning-based risk profiling in breast cancer histopathology and comparison to an established multigene assay.

PURPOSE: To evaluate the Stratipath Breast tool for image-based risk profiling and compare it with an established prognostic multigene assay for risk profiling in a real-world case series of estrogen receptor (ER)-positive and human epidermal growth factor receptor 2 (HER2)-negative early breast cancer patients categorized as intermediate risk based on classic clinicopathological variables and eligible for chemotherapy. METHODS: In a case series comprising 234 invasive ER-positive/HER2-negative tumors, clinicopathological data including Prosigna results and corresponding HE-stained tissue slides were retrieved. The digitized HE slides were analysed by Stratipath Breast. RESULTS: Our findings showed that the Stratipath Breast analysis identified 49.6% of the clinically intermediate tumors as low risk and 50.4% as high risk. The Prosigna assay classified 32.5%, 47.0% and 20.5% tumors as low, intermediate and high risk, respectively. Among Prosigna intermediate-risk tumors, 47.3% were stratified as Stratipath low risk and 52.7% as high risk. In addition, 89.7% of Stratipath low-risk cases were classified as Prosigna low/intermediate risk. The overall agreement between the two tests for low-risk and high-risk groups (N = 124) was 71.0%, with a Cohen's kappa of 0.42. For both risk profiling tests, grade and Ki67 differed significantly between risk groups. CONCLUSION: The results from this clinical evaluation of image-based risk stratification shows a considerable agreement to an established gene expression assay in routine breast pathology.

Development and prognostic validation of a three-level NHG-like deep learning-based model for histological grading of breast cancer.

BACKGROUND: Histological grade is a well-known prognostic factor that is routinely assessed in breast tumours. However, manual assessment of Nottingham Histological Grade (NHG) has high inter-assessor and inter-laboratory variability, causing uncertainty in grade assignments. To address this challenge, we developed and validated a three-level NHG-like deep learning-based histological grade model (predGrade). The primary performance evaluation focuses on prognostic performance. METHODS: This observational study is based on two patient cohorts (SöS-BC-4, N = 2421 (training and internal test); SCAN-B-Lund, N = 1262 (test)) that include routine histological whole-slide images (WSIs) together with patient outcomes. A deep convolutional neural network (CNN) model with an attention mechanism was optimised for the classification of the three-level histological grading (NHG) from haematoxylin and eosin-stained WSIs. The prognostic performance was evaluated by time-to-event analysis of recurrence-free survival and compared to clinical NHG grade assignments in the internal test set as well as in the fully independent external test cohort. RESULTS: We observed effect sizes (hazard ratio) for grade 3 versus 1, for the conventional NHG method (HR = 2.60 (1.18-5.70 95%CI, p-value = 0.017)) and the deep learning model (HR = 2.27, 95%CI 1.07-4.82, p-value = 0.033) on the internal test set after adjusting for established clinicopathological risk factors. In the external test set, the unadjusted HR for clinical NHG 2 versus 1 was estimated to be 2.59 (p-value = 0.004) and clinical NHG 3 versus 1 was estimated to be 3.58 (p-value < 0.001). For predGrade, the unadjusted HR for predGrade 2 versus 1 HR = 2.52 (p-value = 0.030), and 4.07 (p-value = 0.001) for preGrade 3 versus 1 was observed in the independent external test set. In multivariable analysis, HR estimates for neither clinical NHG nor predGrade were found to be significant (p-value > 0.05). We tested for differences in HR estimates between NHG and predGrade in the independent test set and found no significant difference between the two classification models (p-value > 0.05), confirming similar prognostic performance between conventional NHG and predGrade. CONCLUSION: Routine histopathology assessment of NHG has a high degree of inter-assessor variability, motivating the development of model-based decision support to improve reproducibility in histological grading. We found that the proposed model (predGrade) provides a similar prognostic performance as clinical NHG. The results indicate that deep CNN-based models can be applied for breast cancer histological grading.

Transcriptional intra-tumour heterogeneity predicted by deep learning in routine breast histopathology slides provides independent prognostic information.

BACKGROUND: Intra-tumour heterogeneity (ITH) causes diagnostic challenges and increases the risk for disease recurrence. Quantification of ITH is challenging and has not been demonstrated in large studies. It has previously been shown that deep learning can enable spatially resolved prediction of molecular phenotypes from digital histopathology whole slide images (WSIs). Here we propose a novel method (Deep-ITH) to predict and measure ITH, and we evaluate its prognostic performance in breast cancer. METHODS: Deep convolutional neural networks were used to spatially predict gene-expression (PAM50 set) from WSIs. For each predicted transcript, 12 measures of heterogeneity were extracted in the training data set (N = 931). A prognostic score to dichotomise patients into Deep-ITH low- and high-risk groups was established using an elastic-net regularised Cox proportional hazards model (recurrence-free survival). Prognostic performance was evaluated in two independent data sets: SöS-BC-1 (N = 1358) and SCAN-B-Lund (N = 1262). RESULTS: We observed an increase in risk of recurrence in the high-risk group with hazard ratio (HR) 2.11 (95%CI:1.22-3.60; p = 0.007) using nested cross-validation. Subgroup analyses confirmed the prognostic performance in oestrogen receptor (ER)-positive, human epidermal growth factor receptor 2 (HER2)-negative, grade 3, and large tumour subgroups. The prognostic value was confirmed in the independent SöS-BC-1 cohort (HR=1.84; 95%CI:1.03-3.3; p = 3.99 ×10-2). In the other external cohort, significant HR was observed in the subgroup of histological grade 2 patients, as well as in the subgroup of patients with small tumours (<20 mm). CONCLUSION: We developed a novel method for an automated, scalable, and cost-efficient measure of ITH from WSIs that provides independent prognostic value for breast cancer. SIGNIFICANCE: Transcriptional ITH predicted by deep learning models enables prediction of patient survival from routine histopathology WSIs in breast cancer.

Using deep learning to detect patients at risk for prostate cancer despite benign biopsies.

Routine transrectal ultrasound-guided systematic prostate biopsy only samples a small volume of the prostate and tumors between biopsy cores can be missed, leading to low sensitivity to detect clinically relevant prostate cancers (PCa). Deep learning may enable detection of PCa despite benign biopsies. We included 14,354 hematoxylin-eosin stained benign prostate biopsies from 1,508 men in two groups: men without established PCa diagnosis and men with at least one core biopsy diagnosed with PCa. A 10-Convolutional Neural Network ensemble was optimized to distinguish benign biopsies from benign men or patients with PCa. Area under the receiver operating characteristic curve was estimated at 0.739 (bootstrap 95% CI:0.682-0.796) on man level in the held-out test set. At the specificity of 0.90, the model sensitivity was 0.348. The proposed model can detect men with risk of missed PCa and has the potential to reduce false negatives and to indicate men who could benefit from rebiopsies.

Transcriptome-wide prediction of prostate cancer gene expression from histopathology images using co-expression-based convolutional neural networks.

MOTIVATION: Molecular phenotyping by gene expression profiling is central in contemporary cancer research and in molecular diagnostics but remains resource intense to implement. Changes in gene expression occurring in tumours cause morphological changes in tissue, which can be observed on the microscopic level. The relationship between morphological patterns and some of the molecular phenotypes can be exploited to predict molecular phenotypes from routine haematoxylin and eosin-stained whole slide images (WSIs) using convolutional neural networks (CNNs). In this study, we propose a new, computationally efficient approach to model relationships between morphology and gene expression. RESULTS: We conducted the first transcriptome-wide analysis in prostate cancer, using CNNs to predict bulk RNA-sequencing estimates from WSIs for 370 patients from the TCGA PRAD study. Out of 15 586 protein coding transcripts, 6618 had predicted expression significantly associated with RNA-seq estimates (FDR-adjusted P-value <1×10-4) in a cross-validation and 5419 (81.9%) of these associations were subsequently validated in a held-out test set. We furthermore predicted the prognostic cell-cycle progression score directly from WSIs. These findings suggest that contemporary computer vision models offer an inexpensive and scalable solution for prediction of gene expression phenotypes directly from WSIs, providing opportunity for cost-effective large-scale research studies and molecular diagnostics. AVAILABILITY AND IMPLEMENTATION: A self-contained example is available from http://github.com/phiwei/prostate_coexpression. Model predictions and metrics are available from doi.org/10.5281/zenodo.4739097. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Predicting Molecular Phenotypes from Histopathology Images: A Transcriptome-Wide Expression-Morphology Analysis in Breast Cancer.

Molecular profiling is central in cancer precision medicine but remains costly and is based on tumor average profiles. Morphologic patterns observable in histopathology sections from tumors are determined by the underlying molecular phenotype and therefore have the potential to be exploited for prediction of molecular phenotypes. We report here the first transcriptome-wide expression-morphology (EMO) analysis in breast cancer, where individual deep convolutional neural networks were optimized and validated for prediction of mRNA expression in 17,695 genes from hematoxylin and eosin-stained whole slide images. Predicted expressions in 9,334 (52.75%) genes were significantly associated with RNA sequencing estimates. We also demonstrated successful prediction of an mRNA-based proliferation score with established clinical value. The results were validated in independent internal and external test datasets. Predicted spatial intratumor variabilities in expression were validated through spatial transcriptomics profiling. These results suggest that EMO provides a cost-efficient and scalable approach to predict both tumor average and intratumor spatial expression from histopathology images. SIGNIFICANCE: Transcriptome-wide expression morphology deep learning analysis enables prediction of mRNA expression and proliferation markers from routine histopathology whole slide images in breast cancer.

Double-stranded RNA-induced dopaminergic neuronal loss in the substantia nigra in the presence of Mac1 receptor.

BACKGROUND: The underlying mechanism of viral infection as a risk factor for Parkinson's disease (PD), the second most common neurodegenerative disease, remains unclear. OBJECTIVE: We used Mac-1-/- and gp91phox-/- transgene animal models to investigate the mechanisms by which poly I:C, a mimic of virus double-stranded RNA, induces PD neurodegeneration. METHOD: Poly I:C was stereotaxically injected into the substantia nigra (SN) of wild-type (WT), Mac-1-knockout (Mac-1-/-) and gp91 phox-knockout (gp91 phox-/-) mice (10 µg/µl), and nigral dopaminergic neurodegeneration, α-synuclein accumulation and neuroinflammation were evaluated. RESULT: Dopaminergic neurons in the nigra and striatum were markedly reduced in WT mice after administration of poly I:C together with abundant microglial activation in the SN, and the expression of α-synuclein was also elevated. However, these pathological changes were greatly dampened in Mac-1-/- and gp91 phox-/- mice. CONCLUSIONS: Our findings demonstrated that viral infection could result in the activation of microglia as well as NADPH oxidase, which may lead to neuron loss and the development of Parkinson's-like symptoms. Mac-1 is a key receptor during this process.

Benzo(a)pyrene exposure induced neuronal loss, plaque deposition, and cognitive decline in APP/PS1 mice.

BACKGROUND: Exposure to benzo(a)pyrene (BaP) was associated with cognitive impairments and some Alzheimer's disease (AD)-like pathological changes. However, it is largely unknown whether BaP exposure participates in the disease progression of AD. OBJECTIVES: To investigate the effect of BaP exposure on AD progression and its underlying mechanisms. METHODS: BaP or vehicle was administered to 4-month-old APPswe/PS1dE9 transgenic (APP/PS1) mice and wildtype (WT) mice for 2 months. Learning and memory ability and exploratory behaviors were evaluated 1 month after the initiation/termination of BaP exposure. AD-like pathological and biochemical alterations were examined 1 month after 2-month BaP exposure. Levels of soluble beta-amyloid (Aß) oligomers and the number of Aß plaques in the cortex and the hippocampus were quantified. Gene expression profiling was used to evaluate alternation of genes/pathways associated with AD onset and progression. Immunohistochemistry and Western blot were used to demonstrate neuronal loss and neuroinflammation in the cortex and the hippocampus. Treatment of primary neuron-glia cultures with aged Aß (a mixture of monomers, oligomers, and fibrils) and/or BaP was used to investigate mechanisms by which BaP enhanced Aß-induced neurodegeneration. RESULTS: BaP exposure induced progressive decline in spatial learning/memory and exploratory behaviors in APP/PS1 mice and WT mice, and APP/PS1 mice showed severer behavioral deficits than WT mice. Moreover, BaP exposure promoted neuronal loss, Aß burden and Aß plaque formation in APP/PS1 mice, but not in WT mice. Gene expression profiling showed most robust alteration in genes and pathways related to inflammation and immunoregulatory process, Aß secretion and degradation, and synaptic formation in WT and APP/PS1 mice after BaP exposure. Consistently, the cortex and the hippocampus of WT and APP/PS1 mice displayed activation of microglia and astroglia and upregulation of inducible nitric oxide synthase (iNOS), glial fibrillary acidic protein (GFAP), and NADPH oxidase (three widely used neuroinflammatory markers) after BaP exposure. Furthermore, BaP exposure aggravated neurodegeneration induced by aged Aß peptide in primary neuron-glia cultures through enhancing NADPH oxidase-derived oxidative stress. CONCLUSION: Our study showed that chronic exposure to environmental pollutant BaP induced, accelerated, and exacerbated the progression of AD, in which elevated neuroinflammation and NADPH oxidase-derived oxidative insults were key pathogenic events.

Ultrafine carbon particles promote rotenone-induced dopamine neuronal loss through activating microglial NADPH oxidase.

BACKGROUND: Atmospheric ultrafine particles (UFPs) and pesticide rotenone were considered as potential environmental risk factors for Parkinson's disease (PD). However, whether and how UFPs alone and in combination with rotenone affect the pathogenesis of PD remains largely unknown. METHODS: Ultrafine carbon black (ufCB, a surrogate of UFPs) and rotenone were used individually or in combination to determine their roles in chronic dopaminergic (DA) loss in neuron-glia, and neuron-enriched, mix-glia cultures. Immunochemistry using antibody against tyrosine hydroxylase was performed to detect DA neuronal loss. Measurement of extracellular superoxide and intracellular reactive oxygen species (ROS) were performed to examine activation of NADPH oxidase. Genetic deletion and pharmacological inhibition of NADPH oxidase and MAC-1 receptor in microglia were employed to examine their role in DA neuronal loss triggered by ufCB and rotenone. RESULTS: In rodent midbrain neuron-glia cultures, ufCB and rotenone alone caused neuronal death in a dose-dependent manner. In particularly, ufCB at doses of 50 and 100µg/cm2 induced significant loss of DA neurons. More importantly, nontoxic doses of ufCB (10µg/cm2) and rotenone (2nM) induced synergistic toxicity to DA neurons. Microglial activation was essential in this process. Furthermore, superoxide production from microglial NADPH oxidase was critical in ufCB/rotenone-induced neurotoxicity. Studies in mix-glia cultures showed that ufCB treatment activated microglial NADPH oxidase to induce superoxide production. Firstly, ufCB enhanced the expression of NADPH oxidase subunits (gp91phox, p47phox and p40phox); secondly, ufCB was recognized by microglial surface MAC-1 receptor and consequently promoted rotenone-induced p47phox and p67phox translocation assembling active NADPH oxidase. CONCLUSION: ufCB and rotenone worked in synergy to activate NADPH oxidase in microglia, leading to oxidative damage to DA neurons. Our findings delineated the potential role of ultrafine particles alone and in combination with pesticide rotenone in the pathogenesis of PD.

Source attributions of heavy metals in rice plant along highway in Eastern China.

Air and soil pollution from traffic has been considered as a critical issue to crop production and food safety, however, few efforts have been paid on distinguish the source origin of traffic-related contaminants in rice plant along highway. Therefore, we investigated metals (Pb, Cd, Cr, Zn and Cu) concentrations and stable Pb isotope ratios in rice plants exposed and unexposed to highway traffic pollution in Eastern China in 2008. Significant differences in metals concentrations between the exposed and unexposed plants existed in leaf for Pb, Cd and Zn, in stem only for Zn, and in grain for Pb and Cd. About 46% of Pb and 41% of Cd in the grain were attributed to the foliar uptake from atmosphere, and there were no obvious contribution of atmosphere to the accumulations of Cr, Zn and Cu in grain. Except for Zn, all of the heavy metals in stem were attributed to the root uptake from soil, although significant accumulations of Pb and Cd from atmosphere existed in leaf. This indicated that different processes existed in the subsequent translocation of foliar-absorbed heavy metals between rice organs. The distinct separation of stable Pb isotope ratios among rice grain, leaf, stem, soil and vehicle exhaust further provided evidences on the different pathways of heavy metal accumulation in rice plant. These results suggested that further more attentions should be paid to the atmospheric deposition of heavy metals from traffic emission when plan crop layout for food safety along highway.