Predicting nodal metastases in squamous cell carcinoma of the oral tongue using artificial intelligence.

OBJECTIVE: The presence of occult nodal metastases in patients with squamous cell carcinoma (SCC) of the oral tongue has implications for treatment. Upwards of 30% of patients will have occult nodal metastases, yet a significant number of patients undergo unnecessary neck dissection to confirm nodal status. This study sought to predict the presence of nodal metastases in patients with SCC of the oral tongue using a convolutional neural network (CNN) that analyzed visual histopathology from the primary tumor alone. METHODS: Cases of SCC of the oral tongue were identified from the records of a single institution. Only patients with complete pathology data were included in the study. The primary tumors were randomized into 2 groups for training and testing, which was performed at 2 different levels of supervision. Board-certified pathologists annotated each slide. HALO-AI convolutional neural network and image software was used to perform training and testing. Receiver operator characteristic (ROC) curves and the Youden J statistic were used for primary analysis. RESULTS: Eighty-nine cases of SCC of the oral tongue were included in the study. The best performing algorithm had a high level of supervision and a sensitivity of 65% and specificity of 86% when identifying nodal metastases. The area under the curve (AUC) of the ROC curve for this algorithm was 0.729. CONCLUSION: A CNN can produce an algorithm that is able to predict nodal metastases in patients with squamous cell carcinoma of the oral tongue by analyzing the visual histopathology of the primary tumor alone.

The MK2 pathway is linked to G-CSF, cytokine production and metastasis in gastric cancer: a novel intercorrelation analysis approach.

BACKGROUND: Gastric cancer is associated with chronic inflammation, but there is still much to understand about the tumor microenvironment and the underlying tumor-promoting mechanisms. The Map kinase-activated protein kinase 2 (MK2) pathway is a regulator of inflammatory cytokine production that we have been studying in gastrointestinal cancers. Here, we set out to determine the significance of this gene in gastric cancer along with its downstream mediators and if there were differences in the primary tumors with and without metastasis. METHODS: Human gastric cancer tissues with and without metastasis were examined for MK2 expression and cytokine profile in organ culture supernatants. Advanced statistical methods including a lower triangular correlation matrix, novel rooted correlation network, linear and logistic regression modeling along with Kruskal-Wallis testing with Sidak correction for multiple testing were applied to gain understanding of cytokines/chemokines linked to metastasis. RESULTS: The MK2 pathway is strongly linked with metastasis and a panel of cytokines. Gene expression was able to classify gastric cancer metastasis 85.7% of the time. A significant association with a panel of cytokines was found, including G-CSF, GM-CSF, Mip-1ß, IFN-α, MCP-1, IL-1ß, IL-6, and TNF-α. Mip-1ß was found to have the strongest association with MK2 and metastasis after Sidak correction for multiple testing. CONCLUSIONS: MK2 gene expression and a novel associated cytokine panel are linked to gastric cancer metastasis. G-CSF is the strongest cytokine to differentiate between metastasis and non-metastasis patients and had the lowest P value, while Mip-1ß showed the strongest association with MK2 and metastasis after Sidak correction. MK2 and associated cytokines are potential biomarkers for gastric cancer metastasis. The novel intercorrelation analysis approach is a promising method for understanding the complex nature of cytokine/chemokine regulation and links to disease outcome.

Phase II Trial Using a Combination of Oxaliplatin, Capecitabine, and Celecoxib with Concurrent Radiation for Newly Diagnosed Resectable Rectal Cancer.

LESSONS LEARNED: Colorectal cancers exhibit a high level of cyclooxygenase-2 (COX-2) expression with strong preclinical rationale for improved clinical outcomes with COX-2 inhibition. Celecoxib is a COX-2 inhibitor and we have shown that it can be safely combined with capecitabine and oxaliplatin as part of neoadjuvant treatment with radiation therapy (RT) in rectal cancer.There was a significant improvement in skin toxicity with this combination as compared with historical data. Considering the field has moved on to single-agent capecitabine, we believe future trials with capecitabine and celecoxib hold potential. BACKGROUND: Improved survival is seen among patients with rectal cancer who achieve pathologic complete response (pCR) after neoadjuvant therapy. Cyclooxygenase-2 (COX-2) expression is increased in gastrointestinal malignancies and it may serve as a target to enhance pathologic response. A trial combining chemoradiation and COX-2 inhibition was conducted to evaluate the pCR rate, surgical outcomes, survival, and treatment toxicity. METHODS: Patients with resectable (T3-4, N1-2) rectal cancer within 12 cm of the anal verge were included in this phase II clinical trial. The neoadjuvant treatment consisted of capecitabine 850 mg/m2 b.i.d. Monday through Friday for 5 weeks, weekly oxaliplatin 50 mg/m2 intravenous (IV), celecoxib 200 mg b.i.d. daily, along with concurrent 45 gray radiation therapy in 25 fractions. RESULTS: Thirty-two patients were included in the final analysis. The primary endpoint was pCR: 31% (95% confidence interval [CI]: 16%-50%). Secondary endpoints were surgical downstaging (SD): 75% (95% CI: 57%-89%) and sphincter-sparing surgery (SSS): 56% (95% CI: 38%-74%). Common grade >3 toxicities were diarrhea and abnormal liver function tests (9% each). Grade 0 and 1 toxicities included radiation dermatitis (59% and 34%, respectively) and proctitis (63% and 28%, respectively). At 3 years, disease-free survival and overall survival (OS) were 84% (95% CI: 65%-93%) and 94% (95% CI: 77%-98%), respectively. CONCLUSION: Chemoradiation with celecoxib in rectal cancer was well tolerated and demonstrated high rates of pCR, SD, and SSS. Improvement in skin toxicity (34% grade 1 and no grade 3/4) as compared with historical results (43%-78% grade 3/4) seems to be a significant improvement with addition of celecoxib to neoadjuvant chemotherapy.

Blockade of MK2 is protective in inflammation-associated colorectal cancer development.

Chronic inflammation is a risk factor for colorectal cancer. The MAPK-activated protein kinase 2 (MK2) pathway controls multiple cellular processes including p38-dependent inflammation. This is the first study to investigate the role of MK2 in development of colitis-associated colon cancer (CAC). Herein, we demonstrate that MK2(-/-) mice are highly resistant to neoplasm development when exposed to AOM/DSS, while wild type (WT) C57BL/6 develop multiple neoplasms with the same treatment. MK2-specific cytokines IL-1, IL-6 and TNF-α were substantially decreased in AOM/DSS treated MK2(-/-) mouse colon tissues compared with WT mice, which coincided with a marked decrease in macrophage influx. Restoring MK2-competent macrophages by injecting WT bone marrow derived macrophages into MK2(-/-) mice led to partial restoration of inflammatory cytokine production with AOM/DSS treatment; however, macrophages were not sufficient to induce neoplasm development. These results indicate that MK2 functions as an inflammatory regulator to promote colonic neoplasm development and may be a potential target for CAC.

The inflammatory response of human pancreatic cancer samples compared to normal controls.

Pancreatic ductal adenocarcinoma (PDAC) is a poor prognosis cancer with an aggressive growth profile that is often diagnosed at late stage and that has few curative or therapeutic options. PDAC growth has been linked to alterations in the pancreas microbiome, which could include the presence of the fungus Malassezia. We used RNA-sequencing to compare 14 matched tumor and normal (tumor adjacent) pancreatic cancer samples and found Malassezia RNA in both the PDAC and normal tissues. Although the presence of Malassezia was not correlated with tumor growth, a set of immune- and inflammatory-related genes were up-regulated in the PDAC compared to the normal samples, suggesting that they are involved in tumor progression. Gene set enrichment analysis suggests that activation of the complement cascade pathway and inflammation could be involved in pro PDAC growth.

Lymph Node Metastases in Papillary Thyroid Carcinoma can be Predicted by a Convolutional Neural Network: a Multi-Institution Study.

OBJECTIVES: The presence of nodal metastases in patients with papillary thyroid carcinoma (PTC) has both staging and treatment implications. However, lymph nodes are often not removed during thyroidectomy. Prior work has demonstrated the capability of artificial intelligence (AI) to predict the presence of nodal metastases in PTC based on the primary tumor histopathology alone. This study aimed to replicate these results with multi-institutional data. METHODS: Cases of conventional PTC were identified from the records of 2 large academic institutions. Only patients with complete pathology data, including at least 3 sampled lymph nodes, were included in the study. Tumors were designated "positive" if they had at least 5 positive lymph node metastases. First, algorithms were trained separately on each institution's data and tested independently on the other institution's data. Then, the data sets were combined and new algorithms were developed and tested. The primary tumors were randomized into 2 groups, one to train the algorithm and another to test it. A low level of supervision was used to train the algorithm. Board-certified pathologists annotated the slides. HALO-AI convolutional neural network and image software was used to perform training and testing. Receiver operator characteristic curves and the Youden J statistic were used for primary analysis. RESULTS: There were 420 cases used in analyses, 45% of which were negative. The best performing single institution algorithm had an area under the curve (AUC) of 0.64 with a sensitivity and specificity of 65% and 61% respectively, when tested on the other institution's data. The best performing combined institution algorithm had an AUC of 0.84 with a sensitivity and specificity of 68% and 91% respectively. CONCLUSION: A convolutional neural network can produce an accurate and robust algorithm that is capable of predicting nodal metastases from primary PTC histopathology alone even in the setting of multi-institutional data.

G-CSF Is a Novel Mediator of T-Cell Suppression and an Immunotherapeutic Target for Women with Colon Cancer.

PURPOSE: G-CSF enhances colon cancer development. This study defines the prevalence and effects of increased G-CSF signaling in human colon cancers and investigates G-CSF inhibition as an immunotherapeutic strategy against metastatic colon cancer. EXPERIMENTAL DESIGN: Patient samples were used to evaluate G-CSF and G-CSF receptor (G-CSFR) levels by IHC with sera used to measure G-CSF levels. Peripheral blood mononuclear cells were used to assess the rate of G-CSFR+ T cells and IFNγ responses to chronic ex vivo G-CSF. An immunocompetent mouse model of peritoneal metastasis (MC38 cells in C57Bl/6J) was used to determine the effects of G-CSF inhibition (αG-CSF) on survival and the tumor microenvironment (TME) with flow and mass cytometry. RESULTS: In human colon cancer samples, the levels of G-CSF and G-CSFR are higher compared to normal colon tissues from the same patient. High patient serum G-CSF is associated with increases in markers of poor prognosis, (e.g., VEGF, IL6). Circulating T cells from patients express G-CSFR at double the rate of T cells from controls. Prolonged G-CSF exposure decreases T cell IFNγ production. Treatment with αG-CSF shifts both the adaptive and innate compartments of the TME and increases survival (HR, 0.46; P = 0.0237) and tumor T-cell infiltration, activity, and IFNγ response with greater effects in female mice. There is a negative correlation between serum G-CSF levels and tumor-infiltrating T cells in patient samples from women. CONCLUSIONS: These findings support G-CSF as an immunotherapeutic target against colon cancer with greater potential benefit in women.

Ulcerative Colitis: Novel Epithelial Insights Provided by Single Cell RNA Sequencing.

Ulcerative Colitis (UC) is a chronic inflammatory disease of the intestinal tract for which a definitive etiology is yet unknown. Both genetic and environmental factors have been implicated in the development of UC. Recently, single cell RNA sequencing (scRNA-seq) technology revealed cell subpopulations contributing to the pathogenesis of UC and brought new insight into the pathways that connect genome to pathology. This review describes key scRNA-seq findings in two major studies by Broad Institute and University of Oxford, investigating the transcriptomic landscape of epithelial cells in UC. We focus on five major findings: (1) the identification of BEST4 + cells, (2) colonic microfold (M) cells, (3) detailed comparison of the transcriptomes of goblet cells, and (4) colonocytes and (5) stem cells in health and disease. In analyzing the two studies, we identify the commonalities and differences in methodologies, results, and conclusions, offering possible explanations, and validated several cell cluster markers. In systematizing the results, we hope to offer a framework that the broad scientific GI community and GI clinicians can use to replicate or corroborate the extensive new findings that RNA-seq offers.

A Deep Learning Convolutional Neural Network Can Differentiate Between Helicobacter Pylori Gastritis and Autoimmune Gastritis With Results Comparable to Gastrointestinal Pathologists.

CONTEXT.­: Pathology studies using convolutional neural networks (CNNs) have focused on neoplasms, while studies in inflammatory pathology are rare. We previously demonstrated a CNN that differentiates reactive gastropathy, Helicobacter pylori gastritis (HPG), and normal gastric mucosa. OBJECTIVE.­: To determine whether a CNN can differentiate the following 2 gastric inflammatory patterns: autoimmune gastritis (AG) and HPG. DESIGN.­: Gold standard diagnoses were blindly established by 2 gastrointestinal (GI) pathologists. One hundred eighty-seven cases were scanned for analysis by HALO-AI. All levels and tissue fragments per slide were included for analysis. The cases were randomized, 112 (60%; 60 HPG, 52 AG) in the training set and 75 (40%; 40 HPG, 35 AG) in the test set. A HALO-AI correct area distribution (AD) cutoff of 50% or more was required to credit the CNN with the correct diagnosis. The test set was blindly reviewed by pathologists with different levels of GI pathology expertise as follows: 2 GI pathologists, 2 general surgical pathologists, and 2 residents. Each pathologist rendered their preferred diagnosis, HPG or AG. RESULTS.­: At the HALO-AI AD percentage cutoff of 50% or more, the CNN results were 100% concordant with the gold standard diagnoses. On average, autoimmune gastritis cases had 84.7% HALO-AI autoimmune gastritis AD and HP cases had 87.3% HALO-AI HP AD. The GI pathologists, general anatomic pathologists, and residents were on average, 100%, 86%, and 57% concordant with the gold standard diagnoses, respectively. CONCLUSIONS.­: A CNN can distinguish between cases of HPG and autoimmune gastritis with accuracy equal to GI pathologists.

Predicting nodal metastases in papillary thyroid carcinoma using artificial intelligence.

BACKGROUND: The presence of nodal metastases is important in the treatment of papillary thyroid carcinoma (PTC). We present our experience using a convolutional neural network (CNN) to predict the presence of nodal metastases in a series of PTC patients using visual histopathology from the primary tumor alone. METHODS: 174 cases of PTC were evaluated for the presence or absence of lymph metastases. The artificial intelligence (AI) algorithm was trained and tested on its ability to discern between the two groups. RESULTS: The best performing AI algorithm demonstrated a sensitivity and specificity of 94% and 100%, respectively, when identifying nodal metastases. CONCLUSION: A CNN can be used to accurately predict the likelihood of nodal metastases in PTC using visual data from the primary tumor alone.

Comparing Deep Learning and Immunohistochemistry in Determining the Site of Origin for Well-Differentiated Neuroendocrine Tumors.

BACKGROUND: Determining the site of origin for metastatic well-differentiated neuroendocrine tumors (WDNETs) is challenging, and immunohistochemical (IHC) profiles do not always lead to a definitive diagnosis. We sought to determine if a deep-learning convolutional neural network (CNN) could improve upon established IHC profiles in predicting the site of origin in a cohort of WDNETs from the common primary sites. MATERIALS AND METHODS: Hematoxylin and eosin (H&E)-stained tissue microarrays (TMAs) were created using 215 WDNETs arising from the known primary sites. A CNN trained and tested on 60% (n = 130) and 40% (n = 85) of these cases, respectively. One hundred and seventy-nine cases had TMA tissue remaining for the IHC analysis. These cases were stained with IHC markers pPAX8, CDX2, SATB2, and thyroid transcription factor-1 (markers of pancreas/duodenum, ileum/jejunum/duodenum, colorectum/appendix, and lung WDNET sites of origin, respectively). The CNN diagnosis was deemed correct if it designated a majority or plurality of the tumor area as the known site of origin. The IHC diagnosis was deemed correct if the most specific marker for a particular site of origin met an H-score threshold determined by two pathologists. RESULTS: When all cases were considered, the CNN correctly identified the site of origin at a lower rate compared to IHC (72% vs. 82%, respectively). Of the 85 cases in the CNN test set, 66 had sufficient TMA material for IHC stains, thus 66 cases were available for a direct case-by-case comparison of IHC versus CNN. The CNN correctly identified 70% of these cases, while IHC correctly identified 76%, a finding that was not statistically significant (P = 0.56). CONCLUSION: A CNN can identify WDNET site of origin at an accuracy rate close to the current gold standard IHC methods.

A Deep Learning Convolutional Neural Network Can Recognize Common Patterns of Injury in Gastric Pathology.

CONTEXT.­: Most deep learning (DL) studies have focused on neoplastic pathology, with the realm of inflammatory pathology remaining largely untouched. OBJECTIVE.­: To investigate the use of DL for nonneoplastic gastric biopsies. DESIGN.­: Gold standard diagnoses were blindly established by 2 gastrointestinal pathologists. For phase 1, 300 classic cases (100 normal, 100 Helicobacter pylori, 100 reactive gastropathy) that best displayed the desired pathology were scanned and annotated for DL analysis. A total of 70% of the cases for each group were selected for the training set, and 30% were included in the test set. The software assigned colored labels to the test biopsies, which corresponded to the area of the tissue assigned a diagnosis by the DL algorithm, termed area distribution (AD). For Phase 2, an additional 106 consecutive nonclassical gastric biopsies from our archives were tested in the same fashion. RESULTS.­: For Phase 1, receiver operating curves showed near perfect agreement with the gold standard diagnoses at an AD percentage cutoff of 50% for normal (area under the curve [AUC] = 99.7%) and H pylori (AUC = 100%), and 40% for reactive gastropathy (AUC = 99.9%). Sensitivity/specificity pairings were as follows: normal (96.7%, 86.7%), H pylori (100%, 98.3%), and reactive gastropathy (96.7%, 96.7%). For phase 2, receiver operating curves were slightly less discriminatory, with optimal AD cutoffs reduced to 40% across diagnostic groups. The AUCs were 91.9% for normal, 100% for H pylori, and 94.0% for reactive gastropathy. Sensitivity/specificity parings were as follows: normal (73.7%, 79.6%), H pylori (95.7%, 100%), reactive gastropathy (100%, 62.5%). CONCLUSIONS.­: A convolutional neural network can serve as an effective screening tool/diagnostic aid for H pylori gastritis.

A Comprehensive Evaluation of Special AT-rich Sequence-binding Protein 2 (SATB2) Immunohistochemical Staining in Mucinous Tumors From Gastrointestinal and Nongastrointestinal Sites.

Special AT-rich sequence-binding protein 2 (SATB2) is an accurate marker for conventional colorectal carcinoma (CRC), although its sensitivity and specificity in mucinous tumors from the colon and other sites remains unknown. The objective of this study is to evaluate the accuracy of SATB2 expression detected by immunohistochemical assay, as a marker of primary CRC in mucinous adenocarcinomas. SATB2 immunohistochemical stains were performed on whole sections from 63 conventional CRCs (controls), 47 mucinous CRCs (mCRC), and 182 noncolorectal mucinous tumors. SATB2 intensity was scored as 1 to 3 based on the estrogen receptor/progesterone receptor grading system, and the percent positive cells was scored in broad categories as follows: 0 (negative)≤5%, 1=5% to 49%, 2≥50%. An optimal sensitivity/specificity pairing (83% and 95%, respectively) was achieved in the mCRCs when the additive intensity and percent score was ≥3 (ie, intensity score+percent score=total score). Defining this total score (histologic score/"H score") as a "positive" result, the sensitivity of SATB2 for conventional CRC was 98% (62/63) versus 83% (39/47) for mCRCs (P=0.02); whereas 5% (9/182) of all noncolorectal mucinous tumors were considered positive. SATB2 especially demonstrated reduced specificity when applied to mucinous gastroesophageal and breast carcinomas, which showed significant expression in 27% and 9% of cases, respectively. In summary, SATB2 is a less sensitive marker of colorectal origin in mCRC compared with conventional CRC and shows significantly reduced specificity in mucinous gastroesophageal and breast primaries.

SATB2 Versus CDX2: A Battle Royale for Diagnostic Supremacy in Mucinous Tumors.

CONTEXT.­: Metastatic mucinous tumors present a diagnostic challenge for pathologists as tumor histomorphology is often nonspecific and optimal immunoprofiles are still under investigation. OBJECTIVE.­: To present a head-to-head comparison of special AT-rich sequence-binding protein 2 (SATB2) and caudal type homeobox 2 (CDX2) expression in a diverse array of primary mucinous tumors. DESIGN.­: SATB2 and CDX2 immunohistochemical stains were performed on whole sections from 44 mucinous colorectal carcinomas and 175 noncolorectal mucinous tumors. A nuclear scoring system measuring intensity (0-3+) and percentage staining (0 = <5%, 1 = 5%-49%, 2 = ≥50%) was implemented, producing an additive histologic score (H-score). RESULTS.­: SATB2 demonstrated acceptable accuracy at low to moderate expression levels (H-scores of 1-4). With these H-score cutoffs, overall accuracy was greater than 90%. In contrast, CDX2's accuracy rivaled that of SATB2 only at an H-score of 5 (89.0%), as its specificity suffered at lower expression levels (<70.0% at H-scores of 1-4). Using a moderate H-score cutoff of 3 or higher, significant differences for both sensitivity and specificity were identified between SATB2 and CDX2 (P = .01 for sensitivity and P < .001 for specificity), though these stains were near equivalent when each was interpreted as positive at its respective optimal H-score (SATB2 ≥ 3 and CDX2 = 5). CONCLUSIONS.­: SATB2 is a more accurate marker of colorectal origin across a variety of expression levels compared with CDX2 when applied to mucinous tumors from a host of primary sites. However, these stains are near equivalent when each is interpreted at its optimal expression level.

Gross Pathology of Routine Pediatric Tonsillectomy Specimens: Optimizing the Value of Patient Care.

OBJECTIVE: This investigation seeks to evaluate the effect of gross pathologic analysis on our management of patients undergoing routine tonsillectomy and to evaluate charges and reimbursement. STUDY DESIGN: Retrospective chart review from 2005 through 2016. SETTING: Academic medical center. SUBJECTS AND METHODS: Participants were pediatric patients aged 14 years and younger undergoing tonsillectomy for either sleep-disordered breathing or tonsillitis, with tonsillectomy specimens evaluated by pathology, and without any risk factors for pediatric malignancy. Records were reviewed for demographics, surgical indications, and pathology. Abnormal reports prompted an in-depth review of the chart. Charges and reimbursement related to both hospital and professional fees for gross tonsil analysis were evaluated. RESULTS: From 2005 to 2016, 3183 routine pediatric tonsillectomy cases were performed with corresponding specimens that were sent for gross analysis revealing no significant pathologic findings; 1841 were males and 1342 were females. Ten cases underwent microscopy by pathologist order, revealing normal tonsillar tissue. The mean charge per patient for gross analysis was $60.67 if tonsils were together as 1 specimen and $77.67 if tonsils were sent as 2 separate specimens; respective reimbursement amounts were $28.74 and $35.90. CONCLUSIONS: Gross pathologic analysis did not change our management of routine pediatric tonsillectomy patients. Foregoing the practice at our institution would eliminate $19,171.72 to $24,543.72 in charges and $9081.40 to $11,344.40 in reimbursement per year. Eliminating this test would improve the value of patient care by saving health care resources without compromising clinical outcomes.