Democratizing Artificial Intelligence in Anatomic Pathology.

CONTEXT.­: Artificial intelligence is a transforming technology for anatomic pathology. Involvement within the workforce will foster support for algorithm development and implementation. OBJECTIVE.­: To develop a supportive ecosystem that enables pathologists with variable expertise in artificial intelligence to create algorithms in a development environment with seamless transition to a production environment. RESULTS.­: The development team considered internal development and vended solutions. Because of the extended timeline and resource requirements for internal development, a decision was made to use a vended solution. Vendor proposals were solicited and reviewed by pathologists, IT, and security groups. A vendor was selected and pipelines for development and production were established. Proposals for development were solicited from the pathology department. Eighty-four investigators were selected for the initial cohort, receiving training and access to dedicated subject matter experts. A total of 30 of 31 projects progressed through the model development process of annotating, training, and validation. Based on these projects, 15 abstracts were submitted to national meetings. CONCLUSIONS.­: Democratizing artificial intelligence by creating an ecosystem to support pathologists with varying levels of expertise can break down entry barriers, reduce overall cost of algorithm development, improve algorithm quality, and enhance the speed of adoption.

Improved Risk-Stratification Scheme for Mismatch-Repair Proficient Stage II Colorectal Cancers Using the Digital Pathology Biomarker QuantCRC.

PURPOSE: There is a need to improve current risk stratification of stage II colorectal cancer to better inform risk of recurrence and guide adjuvant chemotherapy. We sought to examine whether integration of QuantCRC, a digital pathology biomarker utilizing hematoxylin and eosin-stained slides, provides improved risk stratification over current American Society of Clinical Oncology (ASCO) guidelines. EXPERIMENTAL DESIGN: ASCO and QuantCRC-integrated schemes were applied to a cohort of 398 mismatch-repair proficient (MMRP) stage II colorectal cancers from three large academic medical centers. The ASCO stage II scheme was taken from recent guidelines. The QuantCRC-integrated scheme utilized pT3 versus pT4 and a QuantCRC-derived risk classification. Evaluation of recurrence-free survival (RFS) according to these risk schemes was compared using the log-rank test and HR. RESULTS: Integration of QuantCRC provides improved risk stratification compared with the ASCO scheme for stage II MMRP colorectal cancers. The QuantCRC-integrated scheme placed more stage II tumors in the low-risk group compared with the ASCO scheme (62.5% vs. 42.2%) without compromising excellent 3-year RFS. The QuantCRC-integrated scheme provided larger HR for both intermediate-risk (2.27; 95% CI, 1.32-3.91; P = 0.003) and high-risk (3.27; 95% CI, 1.42-7.55; P = 0.006) groups compared with ASCO intermediate-risk (1.58; 95% CI, 0.87-2.87; P = 0.1) and high-risk (2.24; 95% CI, 1.09-4.62; P = 0.03) groups. The QuantCRC-integrated risk groups remained prognostic in the subgroup of patients that did not receive any adjuvant chemotherapy. CONCLUSIONS: Incorporation of QuantCRC into risk stratification provides a powerful predictor of RFS that has potential to guide subsequent treatment and surveillance for stage II MMRP colorectal cancers.

Development and technical validation of an artificial intelligence model for quantitative analysis of histopathologic features of eosinophilic esophagitis.

Background: In an attempt to provide quantitative, reproducible, and standardized analyses in cases of eosinophilic esophagitis (EoE), we have developed an artificial intelligence (AI) digital pathology model for the evaluation of histologic features in the EoE/esophageal eosinophilia spectrum. Here, we describe the development and technical validation of this novel AI tool. Methods: A total of 10 726 objects and 56.2 mm2 of semantic segmentation areas were annotated on whole-slide images, utilizing a cloud-based, deep learning artificial intelligence platform (Aiforia Technologies, Helsinki, Finland). Our training set consisted of 40 carefully selected digitized esophageal biopsy slides which contained the full spectrum of changes typically seen in the setting of esophageal eosinophilia, ranging from normal mucosa to severe abnormalities with regard to each specific features included in our model. A subset of cases was reserved as independent "test sets" in order to assess the validity of the AI model outside the training set. Five specialized experienced gastrointestinal pathologists scored each feature blindly and independently of each other and of AI model results. Results: The performance of the AI model for all cell type features was similar/non-inferior to that of our group of GI pathologists (F1-scores: 94.5-94.8 for AI vs human and 92.6-96.0 for human vs human). Segmentation area features were rated for accuracy using the following scale: 1. "perfect or nearly perfect" (95%-100%, no significant errors), 2. "very good" (80%-95%, only minor errors), 3. "good" (70%-80%, significant errors but still captures the feature well), 4. "insufficient" (less than 70%, significant errors compromising feature recognition). Rating scores for tissue (1.01), spongiosis (1.15), basal layer (1.05), surface layer (1.04), lamina propria (1.15), and collagen (1.11) were in the "very good" to "perfect or nearly perfect" range, while degranulation (2.23) was rated between "good" and "very good". Conclusion: Our newly developed AI-based tool showed an excellent performance (non-inferior to a group of experienced GI pathologists) for the recognition of various histologic features in the EoE/esophageal mucosal eosinophilia spectrum. This tool represents an important step in creating an accurate and reproducible method for semi-automated quantitative analysis to be used in the evaluation of esophageal biopsies in this clinical context.

Quantitative Pathologic Analysis of Digitized Images of Colorectal Carcinoma Improves Prediction of Recurrence-Free Survival.

BACKGROUND & AIMS: To examine whether quantitative pathologic analysis of digitized hematoxylin and eosin slides of colorectal carcinoma (CRC) correlates with clinicopathologic features, molecular alterations, and prognosis. METHODS: A quantitative segmentation algorithm (QuantCRC) was applied to 6468 digitized hematoxylin and eosin slides of CRCs. Fifteen parameters were recorded from each image and tested for associations with clinicopathologic features and molecular alterations. A prognostic model was developed to predict recurrence-free survival using data from the internal cohort (n = 1928) and validated on an internal test (n = 483) and external cohort (n = 938). RESULTS: There were significant differences in QuantCRC according to stage, histologic subtype, grade, venous/lymphatic/perineural invasion, tumor budding, CD8 immunohistochemistry, mismatch repair status, KRAS mutation, BRAF mutation, and CpG methylation. A prognostic model incorporating stage, mismatch repair, and QuantCRC resulted in a Harrell's concordance (c)-index of 0.714 (95% confidence interval [CI], 0.702-0.724) in the internal test and 0.744 (95% CI, 0.741-0.754) in the external cohort. Removing QuantCRC from the model reduced the c-index to 0.679 (95% CI, 0.673-0.694) in the external cohort. Prognostic risk groups were identified, which provided a hazard ratio of 2.24 (95% CI, 1.33-3.87, P = .004) for low vs high-risk stage III CRCs and 2.36 (95% CI, 1.07-5.20, P = .03) for low vs high-risk stage II CRCs, in the external cohort after adjusting for established risk factors. The predicted median 36-month recurrence rate for high-risk stage III CRCs was 32.7% vs 13.4% for low-risk stage III and 15.8% for high-risk stage II vs 5.4% for low-risk stage II CRCs. CONCLUSIONS: QuantCRC provides a powerful adjunct to routine pathologic reporting of CRC. A prognostic model using QuantCRC improves prediction of recurrence-free survival.

Deep learning assisted quantitative assessment of histopathological markers of Alzheimer's disease and cerebral amyloid angiopathy.

Traditionally, analysis of neuropathological markers in neurodegenerative diseases has relied on visual assessments of stained sections. Resulting semiquantitative scores often vary between individual raters and research centers, limiting statistical approaches. To overcome these issues, we have developed six deep learning-based models, that identify some of the most characteristic markers of Alzheimer's disease (AD) and cerebral amyloid angiopathy (CAA). The deep learning-based models are trained to differentially detect parenchymal amyloid ß (Aß)-plaques, vascular Aß-deposition, iron and calcium deposition, reactive astrocytes, microglia, as well as fibrin extravasation. The models were trained on digitized histopathological slides from brains of patients with AD and CAA, using a workflow that allows neuropathology experts to train convolutional neural networks (CNNs) on a cloud-based graphical interface. Validation of all models indicated a very good to excellent performance compared to three independent expert human raters. Furthermore, the Aß and iron models were consistent with previously acquired semiquantitative scores in the same dataset and allowed the use of more complex statistical approaches. For example, linear mixed effects models could be used to confirm the previously described relationship between leptomeningeal CAA severity and cortical iron accumulation. A similar approach enabled us to explore the association between neuroinflammation and disparate Aß pathologies. The presented workflow is easy for researchers with pathological expertise to implement and is customizable for additional histopathological markers. The implementation of deep learning-assisted analyses of histopathological slides is likely to promote standardization of the assessment of neuropathological markers across research centers, which will allow specific pathophysiological questions in neurodegenerative disease to be addressed in a harmonized way and on a larger scale.

Intermolecular Interactions at the Silica-Liquid Interface Modulate the Fermi Resonance Coupling in Surface Methanol.

The buried solid/liquid interface between hydrophilic fused silica and binary solvent mixtures of acetonitrile (MeCN) and methanol (MeOH) was studied with vibrational sum-frequency generation (vSFG) spectroscopy. Our data showed that at high relative concentrations of methanol, the Fermi resonance peak in the vSFG spectrum is greatly suppressed, and it progressively gains intensity as methanol is diluted with perdeuterated acetonitrile. This phenomenon is quantified by the Fermi resonance coupling coefficient, W, extracted using a two-level model, as well as the experimental intensity ratio, R, of the methyl Fermi resonance band to that of the symmetric stretch. At a 1.0 MeOH mole fraction, W and R values were 10 ± 10 cm-1 and 0.01 ± 0.02, respectively, whereas at a 0.1 mole fraction, W and R increased to 46 ± 4 cm-1 and 0.43 ± 0.16, respectively. This indicates that solvation with acetonitrile effectively tunes the Fermi coupling of methanol vibrations at the silica/liquid interface.

Proof of Concept for a Deep Learning Algorithm for Identification and Quantification of Key Microscopic Features in the Murine Model of DSS-Induced Colitis.

Inflammatory bowel disease (IBD) is a complex disease which leads to life-threatening complications and decreased quality of life. The dextran sulfate sodium (DSS) colitis model in mice is known for rapid screening of candidate compounds. Efficacy assessment in this model relies partly on microscopic semiquantitative scoring, which is time-consuming and subjective. We hypothesized that deep learning artificial intelligence (AI) could be used to identify acute inflammation in H&E-stained sections in a consistent and quantitative manner. Training sets were established using ×20 whole slide images of the entire colon. Supervised training of a Convolutional Neural Network (CNN) was performed using a commercial AI platform to detect the entire colon tissue, the muscle and mucosa layers, and 2 categories within the mucosa (normal and acute inflammation E1). The training sets included slides of naive, vehicle-DSS and cyclosporine A-DSS mice. The trained CNN was able to segment, with a high level of concordance, the different tissue compartments in the 3 groups of mice. The segmented areas were used to determine the ratio of E1-affected mucosa to total mucosa. This proof-of-concept work shows promise to increase efficiency and decrease variability of microscopic scoring of DSS colitis when screening candidate compounds for IBD.

Screening For Bone Marrow Cellularity Changes in Cynomolgus Macaques in Toxicology Safety Studies Using Artificial Intelligence Models.

Many compounds affect the cellularity of hematolymphoid organs including bone marrow. Toxicologic pathologists are tasked with their evaluation as part of safety studies. An artificial intelligence (AI) tool could provide diagnostic support for the pathologist. We looked at the ability of a deep-learning AI model to evaluate whole slide images of macaque sternebrae to identify and enumerate bone marrow hematopoietic cells. The AI model was trained and able to differentiate the hematopoietic cells from the other sternebrae tissues. We compared the model to severity scores in a study with decreased hematopoietic cellularity. The mean cells/mm2 from the model was lower for each increase in severity score. The AI model was trained by 1 pathologist, providing proof of concept that AI model generation can be fast and agile, without the need of a cross disciplinary team and significant effort. We see great potential for the role of AI-based bone marrow screening.

Developing a Qualification and Verification Strategy for Digital Tissue Image Analysis in Toxicological Pathology.

Digital tissue image analysis is a computational method for analyzing whole-slide images and extracting large, complex, and quantitative data sets. However, as with any analysis method, the quality of generated results is dependent on a well-designed quality control system for the entire digital pathology workflow. Such system requires clear procedural controls, appropriate user training, and involvement of specialists to oversee key steps of the workflow. The toxicologic pathologist is responsible for reporting data obtained by digital image analysis and therefore needs to ensure that it is correct. To accomplish that, they must understand the main parameters of the quality control system and should play an integral part in its conception and implementation. This manuscript describes the most common digital tissue image analysis end points and potential sources of analysis errors. In addition, it outlines recommended approaches for ensuring quality and correctness of results for both classical and machine-learning based image analysis solutions, as adapted from a recently proposed Food and Drug Administration regulatory framework for modifications to artificial intelligence/machine learning-based software as a medical device. These approaches are beneficial for any type of toxicopathologic study which uses the described end points and can be adjusted based on the intended use of the image analysis solution.

Tuning the Surface Ordering of Self-Assembled Ionic Surfactants on Semiconducting Single-Walled Carbon Nanotubes: Concentration, Tube Diameter, and Counterions.

We report direct spectroscopic measurements of the macromolecular organization of ionic surfactants on the surface of semiconducting single-walled carbon nanotubes (SWCNTs) within solution-processed thin films. By using vibrational sum frequency generation (VSFG) spectroscopy, sensitive measurements of interfacial surfactant ordering were obtained as a function of surfactant concentration for sodium dodecyl sulfate (SDS)-encapsulated (6,5) and (7,6) SWCNTs with and without excess electrolytes. Anionic surfactants are known to effectively stabilize SWCNTs. The current models suggest a strong influence of the dispersion conditions on the surfactant interfacial macromolecular organization and self-assembly. Direct experimental probes of such an organization using nanotubes of specific chirality are needed to validate the existing models. We found that as the bulk SDS concentration increases near the surfactant critical micelle concentration, the interfacial ordering increased, approaching the formation of cylindrical-like micelles with the nanotube at the core. At the higher surfactant concentrations measured here, the (6,5) SWCNTs produced more ordered structures relative to those with the (7,6) SWCNTs. The relatively larger-diameter (7,6) chiral tubes support enhanced van der Waals (vdW) interactions between the tube carbon surface and the surfactant methylene chain groups that likely increase the density of gauche defects. A new effect arises when the precursor solution is exposed to a small concentration of divalent Ca2+ counterions. We postulate that a salt-bridging configuration on such highly curved surfaces decreases the ordering of interfacial surfactant molecules, resulting in compact, disordered structures. However, this phenomenon was not observed with excess Na+ ions at the same ionic strength. Instead, a modest increase in surfactant ordering was observed with the excess monovalent electrolyte. These results provide new insights for thin film solution processing of vdW nanomaterials and demonstrate that VSFG is a sensitive probe of surfactant organization on nanostructures.

Effectively training the hospice and palliative medicine physician workforce for improved end-of-life health care in the United States.

The widening gap between the demand for palliative care services and the supply of trained palliative care professionals has resulted in considerable end-of-life distress for patients. Without formal training in palliative medicine and end-of-life symptom management, physicians in the United States are less equipped to competently address seriously ill and dying patients' medical, emotional, and spiritual needs. Recent attempts within graduate medical education training deliberately seek to prepare a critical mass of physicians as the new hospice and palliative medicine workforce in the United States. In addition, healthcare reform proposals may re-define the National Health Service Corps (NHSC) post-graduate training over the next five years and the Hospice Medicare Benefit altogether. Healthcare policy options include steady changes at multiple levels of medical training -namely, medical school curriculum mandates, requiring all graduate physician residency training to foster patient-centered communication skills and discussions about advanced directives, and instituting palliative medicine proficiency Continuing Medical Education (CME) requirements for all states' medical licensing boards. Attracting qualified physicians to serve patients at the end of life, innovative medical school loan repayment programs and scholarships will also foster excellence in the field of hospice and palliative medicine. Correcting our current paucity of formal training in palliative medicine better utilizes hospice and restores patients' dignity at the end of life.

Comparison of real-time PCR and hemagglutination assay for quantitation of human polyomavirus JC.

Human polyomavirus JC (JCV), the etiological agent of the disease progressive multifocal leukoencephalopathy (PML) affects immunocompromised patients particularly patients with AIDS. In vitro studies of JCV infection are hampered by the lack of sensitive JCV quantitation tests. Although the hemagglutination (HA) assay has been routinely employed for in vitro quantitation of JCV, its sensitivity is severely limited. We have employed a real-time PCR assay which compares favorably with the HA assay for the in vitro quantitation of JCV. JCV(Mad1), propagated in primary human fetal glial (PHFG) cells in two independent laboratories, was purified and quantitated by the HA assay. Both batches of purified JCV(Mad1) were then serially diluted in Dulbecco's Modified Eagle's Medium to obtain HA titers ranging from 64 to 0.001 HA units (HAU) per 100 microL of virus suspension. DNA was extracted from 100 microL of virus suspension and eluted in 50 microL of buffer, and DNA amplification and quantitation were performed in the Bio-Rad iCycler iQ Multicolor Real-Time PCR Detection System using T-antigen as the target gene. Real-time PCR for quantitation of JCV was sensitive and consistently detected 1.8 x 10(1) copies of JCV DNA, and as low as 0.001 HAU equivalent of JCV. Moreover, there was a strong linear correlation between the HA assay and the DNA copy number of JCV(Mad1). The intra-run and inter-run coefficients of variation for the JCV standard curve were 0.06% to 4.8% and 2.6% to 5.2%, respectively. Based on these data, real-time PCR can replace the less-sensitive HA assay for the reliable detection, quantitation and monitoring of in vitro JCV replication.

A rapid in vitro polyomavirus DNA replication assay.

Traditionally, the Hirt extraction method, a multi-step, labor-intensive and time-consuming procedure, is employed to extract selectively low-molecular weight DNA for polyomavirus DNA replication analyses. DNA replication results obtained with this approach are often inconsistent between replicate samples. To increase the efficiency and reproducibility of the polyomavirus DNA replication assay, we compared the DNA quality and yield using Qiagen Spin Column technology and the Hirt extraction technique. CV-1 cells transfected with SV40 DNA were harvested at days 2, 4, and 6 post-transfection, and DNA was extracted using the Qiagen Spin Column and the Hirt extraction methods. Southern hybridization was performed using a (32)P-labeled linear full-length SV40 DNA probe. Viral DNA replication was quantitated using a BioRad phosphorimager, and results obtained with the two procedures were compared. Southern blot analysis revealed consistent and enhanced SV40 DNA recovery using the Qiagen Spin Column technology, and viral DNA replication over a 6-day period was reproducible among triplicate samples. In addition, Qiagen Spin Column technology reduced the time required to obtain good quality DNA for polyomavirus replication assays from 24 h to less than 3 h. Adoption of this extraction procedure will improve the determination of polyomavirus DNA replication activity, while reducing the investigator's exposure to and disposal of toxic organic compounds.

No clustering for linkage map based on low-copy and undermethylated microsatellites.

Clustering has been reported for conifer genetic maps based on hypomethylated or low-copy molecular markers, resulting in uneven marker distribution. To test this, a framework genetic map was constructed from three types of microsatellites: low-copy, undermethylated, and genomic. These Pinus taeda L. microsatellites were mapped using a three-generation pedigree with 118 progeny. The microsatellites were highly informative; of the 32 markers in intercross configuration, 29 were segregating for three or four alleles in the progeny. The sex-averaged map placed 51 of the 95 markers in 15 linkage groups at LOD > 4.0. No clustering or uneven distribution across the genome was observed. The three types of P. taeda microsatellites were randomly dispersed within each linkage group. The 51 microsatellites covered a map distance of 795 cM, an average distance of 21.8 cM between markers, roughly half of the estimated total map length. The minimum and maximum distances between any two bins was 4.4 and 45.3 cM, respectively. These microsatellites provided anchor points for framework mapping for polymorphism in P. taeda and other closely related hard pines.