Application of deep learning to understand resilience to Alzheimer's disease pathology.

People who have Alzheimer's disease neuropathologic change (ADNC) typically associated with dementia but not the associated cognitive decline can be considered to be "resilient" to the effects of ADNC. We have previously reported lower neocortical levels of hyperphosphorylated tau (pTau) and less limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC) in the resilient participants compared to those with dementia and similar ADNC as determined by current NIA-AA recommendations using traditional semi-quantitative assessments of amyloid ß and pathological tau burden. To better understand differences between AD-dementia and resilient participants, we developed and applied a deep learning approach to analyze the neuropathology of 14 brain donors from the Adult Changes in Thought study, including seven stringently defined resilient participants and seven age-matched AD-dementia controls. We created two novel, fully automated deep learning algorithms to quantify the level of phosphorylated TDP-43 (pTDP-43) and pTau in whole slide imaging. The models performed better than traditional techniques for quantifying pTDP-43 and pTau. The second model was able to segment lesions staining for pTau into neurofibrillary tangles (NFTs) and tau neurites (neuronal processes positive for pTau). Both groups had similar quantities of pTau localizing to neurites, but the pTau burden associated with NFTs in the resilient group was significantly lower compared to the group with dementia. These results validate use of deep learning approaches to quantify clinically relevant microscopic characteristics from neuropathology workups. These results also suggest that the burden of NFTs is more strongly associated with cognitive impairment than the more diffuse neuritic tau commonly seen with tangle pathology and suggest that additional factors may underlie resilience mechanisms defined by traditional means.

Signature maps for automatic identification of prostate cancer from colorimetric analysis of H&E- and IHC-stained histopathological specimens.

Prostate cancer (PCa) is a major cause of cancer death among men. The histopathological examination of post-surgical prostate specimens and manual annotation of PCa not only allow for detailed assessment of disease characteristics and extent, but also supply the ground truth for developing of computer-aided diagnosis (CAD) systems for PCa detection before definitive treatment. As manual cancer annotation is tedious and subjective, there have been a number of publications describing methods for automating the procedure via the analysis of digitized whole-slide images (WSIs). However, these studies have focused only on the analysis of WSIs stained with hematoxylin and eosin (H&E), even though there is additional information that could be obtained from immunohistochemical (IHC) staining. In this work, we propose a framework for automating the annotation of PCa that is based on automated colorimetric analysis of both H&E and IHC WSIs stained with a triple-antibody cocktail against high-molecular weight cytokeratin (HMWCK), p63, and α-methylacyl CoA racemase (AMACR). The analysis outputs were then used to train a regression model to estimate the distribution of cancerous epithelium within slides. The approach yielded an AUC of 0.951, sensitivity of 87.1%, and specificity of 90.7% as compared to slide-level annotations, and generalized well to cancers of all grades.

Safe and Effective Sarcoma Therapy through Bispecific Targeting of EGFR and uPAR.

Sarcomas differ from carcinomas in their mesenchymal origin. Therapeutic advancements have come slowly, so alternative drugs and models are urgently needed. These studies report a new drug for sarcomas that simultaneously targets both tumor and tumor neovasculature. eBAT is a bispecific angiotoxin consisting of truncated, deimmunized Pseudomonas exotoxin fused to EGF and the amino terminal fragment of urokinase. Here, we study the drug in an in vivo "ontarget" companion dog trial as eBAT effectively kills canine hemangiosarcoma and human sarcoma cells in vitro We reasoned the model has value due to the common occurrence of spontaneous sarcomas in dogs and a limited lifespan allowing for rapid accrual and data collection. Splenectomized dogs with minimal residual disease were given one cycle of eBAT followed by adjuvant doxorubicin in an adaptive dose-finding, phase I-II study of 23 dogs with spontaneous, stage I-II, splenic hemangiosarcoma. eBAT improved 6-month survival from <40% in a comparison population to approximately 70% in dogs treated at a biologically active dose (50 µg/kg). Six dogs were long-term survivors, living >450 days. eBAT abated expected toxicity associated with EGFR targeting, a finding supported by mouse studies. Urokinase plasminogen activator receptor and EGFR are targets for human sarcomas, so thorough evaluation is crucial for validation of the dog model. Thus, we validated these markers for human sarcoma targeting in the study of 212 human and 97 canine sarcoma samples. Our results support further translation of eBAT for human patients with sarcomas and perhaps other EGFR-expressing malignancies. Mol Cancer Ther; 16(5); 956-65. ©2017 AACR.

NDER: A novel web application using annotated whole slide images for rapid improvements in human pattern recognition.

CONTEXT: Whole-slide images (WSIs) present a rich source of information for education, training, and quality assurance. However, they are often used in a fashion similar to glass slides rather than in novel ways that leverage the advantages of WSI. We have created a pipeline to transform annotated WSI into pattern recognition training, and quality assurance web application called novel diagnostic electronic resource (NDER). AIMS: Create an efficient workflow for extracting annotated WSI for use by NDER, an attractive web application that provides high-throughput training. MATERIALS AND METHODS: WSI were annotated by a resident and classified into five categories. Two methods of extracting images and creating image databases were compared. Extraction Method 1: Manual extraction of still images and validation of each image by four breast pathologists. Extraction Method 2: Validation of annotated regions on the WSI by a single experienced breast pathologist and automated extraction of still images tagged by diagnosis. The extracted still images were used by NDER. NDER briefly displays an image, requires users to classify the image after time has expired, then gives users immediate feedback. RESULTS: The NDER workflow is efficient: annotation of a WSI requires 5 min and validation by an expert pathologist requires An additional one to 2 min. The pipeline is highly automated, with only annotation and validation requiring human input. NDER effectively displays hundreds of high-quality, high-resolution images and provides immediate feedback to users during a 30 min session. CONCLUSIONS: NDER efficiently uses annotated WSI to rapidly increase pattern recognition and evaluate for diagnostic proficiency.

Quantitative comparison and reproducibility of pathologist scoring and digital image analysis of estrogen receptor ß2 immunohistochemistry in prostate cancer.

BACKGROUND: Digital image analysis offers advantages over traditional pathologist visual scoring of immunohistochemistry, although few studies examining the correlation and reproducibility of these methods have been performed in prostate cancer. We evaluated the correlation between digital image analysis (continuous variable data) and pathologist visual scoring (quasi-continuous variable data), reproducibility of each method, and association of digital image analysis methods with outcomes using prostate cancer tissue microarrays (TMAs) stained for estrogen receptor-ß2 (ERß2). METHODS: Prostate cancer TMAs were digitized and evaluated by pathologist visual scoring versus digital image analysis for ERß2 staining within tumor epithelium. Two independent analysis runs were performed to evaluate reproducibility. Image analysis data were evaluated for associations with recurrence-free survival and disease specific survival following radical prostatectomy. RESULTS: We observed weak/moderate Spearman correlation between digital image analysis and pathologist visual scores of tumor nuclei (Analysis Run A: 0.42, Analysis Run B: 0.41), and moderate/strong correlation between digital image analysis and pathologist visual scores of tumor cytoplasm (Analysis Run A: 0.70, Analysis Run B: 0.69). For the reproducibility analysis, there was high Spearman correlation between pathologist visual scores generated for individual TMA spots across Analysis Runs A and B (Nuclei: 0.84, Cytoplasm: 0.83), and very high correlation between digital image analysis for individual TMA spots across Analysis Runs A and B (Nuclei: 0.99, Cytoplasm: 0.99). Further, ERß2 staining was significantly associated with increased risk of prostate cancer-specific mortality (PCSM) when quantified by cytoplasmic digital image analysis (HR 2.16, 95 % CI 1.02-4.57, p = 0.045), nuclear image analysis (HR 2.67, 95 % CI 1.20-5.96, p = 0.016), and total malignant epithelial area analysis (HR 5.10, 95 % CI 1.70-15.34, p = 0.004). After adjusting for clinicopathologic factors, only total malignant epithelial area ERß2 staining was significantly associated with PCSM (HR 4.08, 95 % CI 1.37-12.15, p = 0.012). CONCLUSIONS: Digital methods of immunohistochemical quantification are more reproducible than pathologist visual scoring in prostate cancer, suggesting that digital methods are preferable and especially warranted for studies involving large sample sizes.

Detection of Prostate Cancer: Quantitative Multiparametric MR Imaging Models Developed Using Registered Correlative Histopathology.

Purpose To develop multiparametric magnetic resonance (MR) imaging models to generate a quantitative, user-independent, voxel-wise composite biomarker score (CBS) for detection of prostate cancer by using coregistered correlative histopathologic results, and to compare performance of CBS-based detection with that of single quantitative MR imaging parameters. Materials and Methods Institutional review board approval and informed consent were obtained. Patients with a diagnosis of prostate cancer underwent multiparametric MR imaging before surgery for treatment. All MR imaging voxels in the prostate were classified as cancer or noncancer on the basis of coregistered histopathologic data. Predictive models were developed by using more than one quantitative MR imaging parameter to generate CBS maps. Model development and evaluation of quantitative MR imaging parameters and CBS were performed separately for the peripheral zone and the whole gland. Model accuracy was evaluated by using the area under the receiver operating characteristic curve (AUC), and confidence intervals were calculated with the bootstrap procedure. The improvement in classification accuracy was evaluated by comparing the AUC for the multiparametric model and the single best-performing quantitative MR imaging parameter at the individual level and in aggregate. Results Quantitative T2, apparent diffusion coefficient (ADC), volume transfer constant (K(trans)), reflux rate constant (kep), and area under the gadolinium concentration curve at 90 seconds (AUGC90) were significantly different between cancer and noncancer voxels (P < .001), with ADC showing the best accuracy (peripheral zone AUC, 0.82; whole gland AUC, 0.74). Four-parameter models demonstrated the best performance in both the peripheral zone (AUC, 0.85; P = .010 vs ADC alone) and whole gland (AUC, 0.77; P = .043 vs ADC alone). Individual-level analysis showed statistically significant improvement in AUC in 82% (23 of 28) and 71% (24 of 34) of patients with peripheral-zone and whole-gland models, respectively, compared with ADC alone. Model-based CBS maps for cancer detection showed improved visualization of cancer location and extent. Conclusion Quantitative multiparametric MR imaging models developed by using coregistered correlative histopathologic data yielded a voxel-wise CBS that outperformed single quantitative MR imaging parameters for detection of prostate cancer, especially when the models were assessed at the individual level. (©) RSNA, 2016 Online supplemental material is available for this article.

Evaluation of protein biomarkers of prostate cancer aggressiveness.

BACKGROUND: Prognostic multibiomarker signatures in prostate cancer (PCa) may improve patient management and provide a bridge for developing novel therapeutics and imaging methods. Our objective was to evaluate the association between expression of 33 candidate protein biomarkers and time to biochemical failure (BF) after prostatectomy. METHODS: PCa tissue microarrays were constructed representing 160 patients for whom clinicopathologic features and follow-up data after surgery were available. Immunohistochemistry for each of 33 proteins was quantified using automated digital pathology techniques. Relationships between clinicopathologic features, staining intensity, and time to BF were assessed. Predictive modeling using multiple imputed datasets was performed to identify the top biomarker candidates. RESULTS: In univariate analyses, lymph node positivity, surgical margin positivity, non-localized tumor, age at prostatectomy, and biomarkers CCND1, HMMR, IGF1, MKI67, SIAH2, and SMAD4 in malignant epithelium were significantly associated with time to BF. HMMR, IGF1, and SMAD4 remained significantly associated with BF after adjusting for clinicopathologic features while additional associations were observed for HOXC6 and MAP4K4 following adjustment. In multibiomarker predictive models, 3 proteins including HMMR, SIAH2, and SMAD4 were consistently represented among the top 2, 3, 4, and 5 most predictive biomarkers, and a signature comprised of these proteins best predicted BF at 3 and 5 years. CONCLUSIONS: This study provides rationale for investigation of HMMR, HOXC6, IGF1, MAP4K4, SIAH2, and SMAD4 as biomarkers of PCa aggressiveness in larger cohorts.

Elevated hyaluronan and hyaluronan-mediated motility receptor are associated with biochemical failure in patients with intermediate-grade prostate tumors.

BACKGROUND: The clinical course of prostate cancer (PCa) measured by biochemical failure (BF) after prostatectomy remains unpredictable in many patients, particularly in intermediate Gleason score (GS) 7 tumors, suggesting that identification of molecular mechanisms associated with aggressive PCa biology may be exploited for improved prognostication or therapy. Hyaluronan (HA) is a high molecular weight polyanionic carbohydrate produced by synthases (HAS1 through HAS3) and fragmented by oxidative/nitrosative stress and hyaluronidases (HYAL1 through HYAL4, SPAM1) common in PCa microenvironments. HA and HA fragments interact with receptors CD44 and hyaluronan-mediated motility receptor (HMMR), resulting in increased tumor aggressiveness in experimental PCa models. This study evaluated the association of HA-related molecules with BF after prostatectomy in GS7 tumors. METHODS: Tissue microarrays were constructed from a 96-patient cohort. HA histochemistry and HAS2, HYAL1, CD44, CD44v6, and HMMR immunohistochemistry were quantified using digital pathology techniques. RESULTS: HA in tumor-associated stroma and HMMR in malignant epithelium were significantly and marginally significantly associated with time to BF in univariate analysis, respectively. After adjusting for clinicopathologic features, both HA in tumor-associated stroma and HMMR in malignant epithelium were significantly associated with time to BF. Although not significantly associated with BF, HAS2 and HYAL1 positively correlated with HMMR in malignant epithelium. Cell culture assays demonstrated that HMMR bound native and fragmented HA, promoted HA uptake, and was required for a promigratory response to fragmented HA. CONCLUSIONS: HA and HMMR are factors associated with time to BF in GS7 tumors, suggesting that increased HA synthesis and fragmentation within the tumor microenvironment stimulates aggressive PCa behavior through HA-HMMR signaling.

Expression of FAP, ADAM12, WISP1, and SOX11 is heterogeneous in aggressive fibromatosis and spatially relates to the histologic features of tumor activity.

Aggressive fibromatosis (AF) represents a group of tumors with a variable and unpredictable clinical course, characterized by a monoclonal proliferation of myofibroblastic cells. The optimal treatment for AF remains unclear. Identification and validation of genes whose expression patterns are associated with AF may elucidate biological mechanisms in AF, and aid treatment selection. This study was designed to examine the protein expression by immunohistochemistry (IHC) of four genes, ADAM12, FAP, SOX11, and WISP1, that were found in an earlier study to be uniquely overexpressed in AF compared with normal tissues. Digital image analysis was performed to evaluate inter- and intratumor heterogeneity, and correlate protein expression with histologic features, including a histopathologic assessment of tumor activity, defined by nuclear chromatin density ratio (CDR). AF tumors exhibited marked inter- and intratumor histologic heterogeneity. Pathologic assessment of tumor activity and digital assessment of average nuclear size and CDR were all significantly correlated. IHC revealed protein expression of all four genes. IHC staining for ADAM12, FAP, and WISP1 correlated with CDR and was higher, whereas SOX11 staining was lower in tumors with earlier recurrence following excision. All four proteins were expressed, and the regional variation in tumor activity within and among AF cases was demonstrated. A spatial correlation between protein expression and nuclear morphology was observed. IHC also correlated with the probability of recurrence following excision. These proteins may be involved in AF pathogenesis and the corresponding pathways could serve as potential targets of therapy.

Time trends in fungal infections as a cause of death in hematopoietic stem cell transplant recipients: an autopsy study.

Infectious complications remain an important cause of death in hematopoietic stem cell transplant (HSCT) recipients. We undertook a 20-year (1988-2007) retrospective review of all autopsies performed on HSCT recipients in our institution, with emphasis on infections, especially fungal infections, as the cause of death. Of the 2,943 autopsies performed in our institution from 1988 to 2007, 395 (13.4%) involved HSCT recipients (117 pediatric; 278 adult). Of the patients, 298 had received allogeneic, 46 autologous, 41 umbilical cord blood, and 3 autologous plus allogeneic types, and 7 were unknown HSCT types. The most common causes of death were pulmonary complications, occurring in 247 (62.5%) of 395 cases. In 178 cases (45.1%), microorganisms (viral, bacterial, and/or fungal) were documented at autopsy in one or more organs and contributed to the cause of death. Fungal infections were found in 23.5% of cases, but their frequency as a cause of death decreased throughout this study, from 30.3% in the 1988-1992 period to 10.9% in the 2003-2007 period.

Visual estimates of nucleus-to-nucleus ratios: can we trust our eyes to use the Bethesda ASCUS and LSIL size criteria?

BACKGROUND: Apart from several subjective criteria, ASCUS and LSIL are defined by nuclear enlargement of 2.5x to 3x and > or = 3x the area of a normal intermediate squamous cell nucleus, respectively. The aim of this study was to assess the ability of observers with various degrees of experience to estimate nuclear area ratios. METHODS: Forty-five participants (5 anatomic pathologists, 5 cytopathologists, 2 cytopathology fellows, 16 pathology residents, 8 cytotechnologists, 6 medical students, and 3 cytopreparatory staff members) judged the area ratios of pairs of squamous cell nuclei imaged at x100 objective magnification. For Test 1 (T1), participants chose between 5 preset area ratios (1.0x to 1.4x, 1.5x to 1.9x, 2.0x to 2.4x, 2.5x to 2.9x, 3.0x to 3.4x) for 15 pairs of cells with ratios falling in the middle of these intervals. One week after T1, an instructional tutorial was given and T1 was repeated in different order of cell pairs as Test 2 (T2). RESULTS: The kappa values for all participants were 0.30 for T1 and 0.39 for T2. Accurate responses were given in 50.5% in T1 and 53.5% in T2. Both T1 and T2 responses failed to differentiate between area ratios corresponding to ASCUS and LSIL. There were no differences in accuracy according to sex, experience, or expertise in cytopathology. CONCLUSIONS: The overall accuracy of participants in estimating nuclear area ratios was low, especially for ratios in the range of ASCUS versus LSIL, with only minimal improvement after an instructional tutorial. Nuclear area ratio estimation may, therefore, contribute to the well-known substantial variability in ASCUS rates and ASC/squamous intraepithelial lesion ratios.