Does the use of intraoperative frozen section of bone marrow from the cut end of the mandible help assess the adequacy of bone margins following mandibulectomy for oral cancer?

BACKGROUND

The adequacy of the cut end of the mandible following a segmental mandibulectomy done for oral cancer intraoperatively is at times assessed using a frozen section (FS) of the bone marrow (BM) at the cut ends. The study aimed to assess its utility to guide the intraoperative decision on the adequacy of bony margins.

MATERIALS AND METHODS

All patients with oral squamous cell carcinoma (OSCC) who underwent segmental mandibulectomy from January 2012 to December 2021 at our institute and for whom intraoperative FS of BM was utilized were included. We analyzed the sensitivity, specificity, and positive and negative predictive value (PPV, NPV) of this in predicting positive bone margins.

RESULTS

A total of 457 patients were included in the study. The majority of the cases were per premium cases (n = 372, 81.4%). The median age of the cohort was 52 years (range: 22-80 years). Most patients had T4 disease (n = 406, 88.8%). On FS, BM was positive in only 18 patients (3.9%) for whom the bone margin was revised. BM biopsy report in the final histopathology was positive in 12 patients (2.2%). The sensitivity, specificity PPV, and NPV were 52.3%, 98.65%, 64.7%, and 97.7% respectively. No factors predicting BM positivity on FS could be identified in this cohort.

CONCLUSIONS

The BM FS was positive in only a small percentage of patients, and it helped in reducing the bone margin positivity rate from 3.9% to 2.2% only. Hence the intraoperative BM FS seems to have limited utility as seen from our study.

Association of Cutibacterium acnes with human thyroid cancer

Introduction

The diverse subtypes of thyroid carcinoma have distinct clinical outcomes despite a comparable spectrum of underlying genetic alterations. Beyond genetic alterations, sparse efforts have been made to characterize the microbes associated with thyroid cancer. In this study, we examine the microbial profile of thyroid cancer.

Methods

We sequenced the whole transcriptome of 70 thyroid cancers (40 papillary and 30 anaplastic). Using Infectious Pathogen Detector IPD 2.0, we analysed the relative abundance of 1060 microbes across 70 tumours from patients with thyroid cancer against 118 tumour samples from patients with breast, cervical, colorectal, and tongue cancer.

Results

Our analysis reveals a significant prevalence of Cutibacterium acnes in 58.6% thyroid cancer samples compared to other cancer types (p=0.00038). Immune cell fraction analysis between thyroid cancer samples with high and low Cutibacterium loads identify enrichment of immunosuppressive cells, including Tregs (p=0.015), and other anti-inflammatory cytokines in the tumour microenvironment, suggesting an immune evasion/immunosuppression milieu is associated with the infection. A higher burden of Cutibacterium acnes was also found to be associated with poor survival defining a distinct sub-group of thyroid cancer.

Conclusion

Cutibacterium acnes is associated with immune suppression and poor prognosis in a subpopulation of thyroid cancer. This study may help design novel therapeutic measures involving appropriate antibiotics to manage the disease better.

Hepatocellular carcinoma (HCC) masquerading as hilar cholangiocarcinoma: An unusual presentation of jaundice

Jaundice usually occurs in the late stages of hepatocellular carcinoma (HCC). Obstructive jaundice is rarely seen as an initial presentation of HCC, as opposed to cholangiocarcinoma. Various causes of obstructive jaundice in these cases also known as "Icteric HCC" have been described such as tumour thrombi, compression, infiltration or tumours arising from native hepatocytes in the bile duct. We present a case of 74-year-old gentleman with "Icteric HCC" that clinically and radiologically mimicked cholangiocarcinoma for which the patient underwent left hepatectomy with Roux-en-Y hepaticojejunostomy. Histopathology revealed dilated large duct with polygonal sheets of cells of hepatoid morphology which stained diffusely positive for both glypican 3 and Hep-par 1. The epicentre was in the left hepatic duct with no discernible liver lesion and the tumour probably originated from the ectopic hepatocytes within the biliary duct The patient was disease free at 1.5 years of follow up. In conclusion, HCC should be a differential for obstructive jaundice. Patients with such "Icteric HCC" benefit from surgical resection with favourable outcomes. The prognosis in such patients is better than in patients of HCC with jaundice due to hepatic insufficiency.

HRAS-mutated primary thyroid malignant melanoma or medullary thyroid carcinoma with melanocytic dedifferentiation? A singular case with an ontogeny-phylogeny quandary

Melanotic pigment in the thyroid is practically synonymous with chronic minocycline therapy and rare cases of melanotic medullary thyroid carcinoma. However, primary melanoma of the thyroid has not been reported yet. We report a rare case of a 25-year-old male with a locally aggressive thyroid mass and distant metastases at presentation. Radiologically, a 8.3×7.6-cm nodule was identified in the right thyroid lobe. Fine-needle aspiration cytology (FNAC) showed discohesive atypical plasmacytoid cells with prominent nucleoli and no cytoplasmic pigmentation. Serum calcitonin levels were normal. A trucut biopsy showed a malignant tumor with a similar cytomorphology, including marked nuclear pleomorphism. In addition, intracytoplasmic melanin was seen in <1% of cells. Tumor cells were immunonegative for AE1/AE3, TTF1, synaptophysin, and chromogranin while positive for SOX10, S100P, HMB45, and Melan A, confirming the diagnosis of malignant melanoma, without any detectable MTC component in the biopsy. An HRAS G13R mutation was detected on NGS, which, intriguingly, is a known mutation in MTC, and exceedingly rare in melanocytic lesions. No other clinically or radiologically apparent primary lesion was identified elsewhere in the patient. The unusual histology and hitherto unreported molecular findings make this case of primary thyroid melanocytic neoplasm worth reporting. Abstruse origin of melanoma cells in the thyroid gland with molecular signature suggestive of MTC in our case raises a nomenclature and management conundrum, prompting us to revisit the "ontogeny recapitulates phylogeny" theory.

Multiscale deep learning framework captures systemic immune features in lymph nodes predictive of triple negative breast cancer outcome in large-scale studies

The suggestion that the systemic immune response in lymph nodes (LNs) conveys prognostic value for triple-negative breast cancer (TNBC) patients has not previously been investigated in large cohorts. We used a deep learning (DL) framework to quantify morphological features in haematoxylin and eosin-stained LNs on digitised whole slide images. From 345 breast cancer patients, 5,228 axillary LNs, cancer-free and involved, were assessed. Generalisable multiscale DL frameworks were developed to capture and quantify germinal centres (GCs) and sinuses. Cox regression proportional hazard models tested the association between smuLymphNet-captured GC and sinus quantifications and distant metastasis-free survival (DMFS). smuLymphNet achieved a Dice coefficient of 0.86 and 0.74 for capturing GCs and sinuses, respectively, and was comparable to an interpathologist Dice coefficient of 0.66 (GC) and 0.60 (sinus). smuLymphNet-captured sinuses were increased in LNs harbouring GCs (p < 0.001). smuLymphNet-captured GCs retained clinical relevance in LN-positive TNBC patients whose cancer-free LNs had on average ≥2 GCs, had longer DMFS (hazard ratio [HR] = 0.28, p = 0.02) and extended GCs' prognostic value to LN-negative TNBC patients (HR = 0.14, p = 0.002). Enlarged smuLymphNet-captured sinuses in involved LNs were associated with superior DMFS in LN-positive TNBC patients in a cohort from Guy's Hospital (multivariate HR = 0.39, p = 0.039) and with distant recurrence-free survival in 95 LN-positive TNBC patients of the Dutch-N4plus trial (HR = 0.44, p = 0.024). Heuristic scoring of subcapsular sinuses in LNs of LN-positive Tianjin TNBC patients (n = 85) cross-validated the association of enlarged sinuses with shorter DMFS (involved LNs: HR = 0.33, p = 0.029 and cancer-free LNs: HR = 0.21 p = 0.01). Morphological LN features reflective of cancer-associated responses are robustly quantifiable by smuLymphNet. Our findings further strengthen the value of assessment of LN properties beyond the detection of metastatic deposits for prognostication of TNBC patients. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Computational pathology in cancer diagnosis, prognosis, and prediction - present day and prospects

Computational pathology refers to applying deep learning techniques and algorithms to analyse and interpret histopathology images. Advances in artificial intelligence (AI) have led to an explosion in innovation in computational pathology, ranging from the prospect of automation of routine diagnostic tasks to the discovery of new prognostic and predictive biomarkers from tissue morphology. Despite the promising potential of computational pathology, its integration in clinical settings has been limited by a range of obstacles including operational, technical, regulatory, ethical, financial, and cultural challenges. Here, we focus on the pathologists' perspective of computational pathology: we map its current translational research landscape, evaluate its clinical utility, and address the more common challenges slowing clinical adoption and implementation. We conclude by describing contemporary approaches to drive forward these techniques. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Imaging of Neck Nodes in Head and Neck Cancers - a Comprehensive Update

Cervical lymph node metastases from head and neck squamous cell cancers significantly reduce disease-free survival and worsen overall prognosis and, hence, deserve more aggressive management and follow-up. As per the eighth edition of the American Joint Committee on Cancer staging manual, extranodal extension, especially in human papillomavirus-negative cancers, has been incorporated in staging as it is important in deciding management and significantly impacts the outcome of head and neck squamous cell cancer. Lymph node imaging with various radiological modalities, including ultrasound, computed tomography and magnetic resonance imaging, has been widely used, not only to demonstrate nodal involvement but also for guided histopathological evaluation and therapeutic intervention. Computed tomography and magnetic resonance imaging, together with positron emission tomography, are used widely for the follow-up of treated patients. Finally, there is an emerging role for artificial intelligence in neck node imaging that has shown promising results, increasing the accuracy of detection of nodal involvement, especially normal-appearing nodes. The aim of this review is to provide a comprehensive overview of the diagnosis and management of involved neck nodes with a focus on sentinel node anatomy, pathogenesis, imaging correlates (including radiogenomics and artificial intelligence) and the role of image-guided interventions.

Efficient quality control of whole slide pathology images with human-in-the-loop training

Histopathology whole slide images (WSIs) are being widely used to develop deep learning-based diagnostic solutions, especially for precision oncology. Most of these diagnostic softwares are vulnerable to biases and impurities in the training and test data which can lead to inaccurate diagnoses. For instance, WSIs contain multiple types of tissue regions, at least some of which might not be relevant to the diagnosis. We introduce HistoROI, a robust yet lightweight deep learning-based classifier to segregate WSI into 6 broad tissue regions-epithelium, stroma, lymphocytes, adipose, artifacts, and miscellaneous. HistoROI is trained using a novel human in-the-loop and active learning paradigm that ensures variations in training data for labeling efficient generalization. HistoROI consistently performs well across multiple organs, despite being trained on only a single dataset, demonstrating strong generalization. Further, we have examined the utility of HistoROI in improving the performance of downstream deep learning-based tasks using the CAMELYON breast cancer lymph node and TCGA lung cancer datasets. For the former dataset, the area under the receiver operating characteristic curve (AUC) for metastasis versus normal tissue of a neural network trained using weakly supervised learning increased from 0.88 to 0.92 by filtering the data using HistoROI. Similarly, the AUC increased from 0.88 to 0.93 for the classification between adenocarcinoma and squamous cell carcinoma on the lung cancer dataset. We also found that the performance of the HistoROI improves upon HistoQC for artifact detection on a test dataset of 93 annotated WSIs. The limitations of the proposed model are analyzed, and potential extensions are also discussed.

Cautious Artificial Intelligence Improves Outcomes and Trust by Flagging Outlier Cases

PURPOSE

Artificial intelligence (AI) models for medical image diagnosis are often trained and validated on curated data. However, in a clinical setting, images that are outliers with respect to the training data, such as those representing rare disease conditions or acquired using a slightly different setup, can lead to wrong decisions. It is not practical to expect clinicians to be trained to discount results for such outlier images. Toward clinical deployment, we have designed a method to train cautious AI that can automatically flag outlier cases.

MATERIALS AND METHODS

Our method-ClassClust-forms tight clusters of training images using supervised contrastive learning, which helps it identify outliers during testing. We compared ClassClust's ability to detect outliers with three competing methods on four publicly available data sets covering pathology, dermatoscopy, and radiology. We held out certain diseases, artifacts, and types of images from training data and examined the ability of various models to detect these as outliers during testing. We compared the decision accuracy of the models on held-out nonoutlier images also. We visualized the regions of the images that the models used for their decisions.

RESULTS

Area under receiver operating characteristic curve for outlier detection was consistently higher using ClassClust compared with the previous methods. Average accuracy on held-out nonoutlier images was also higher, and the visualizations of image regions were more informative using ClassClust.

CONCLUSION

The ability to flag outlier test cases need not be at odds with the ability to accurately classify nonoutliers in AI models. Although the latter capability has received research and regulatory attention, AI models for clinical deployment should possess the former as well.

Prevention of carcinogen-induced oral cancers by polymeric black tea polyphenols via modulation of EGFR-Akt-mTOR pathway

The overexpression of Epidermal Growth Factor Receptor (EGFR) and dysregulation of its downstream effector pathways are important molecular hallmarks of oral cancers. Present study investigates the chemopreventive potential of polymeric black tea polyphenols (PBPs)/thearubigins (TRs) in the hamster model of oral carcinogenesis as well as determine the effect of PBPs on EGFR and the molecular players in the EGFR pathway. In dose-dependent manner, pre and concurrent treatment with PBPs (1.5%, 5%, 10%) decreased the number and volume of macroscopic tumors as well as the number and area of microscopic lesions. Interestingly, at 10% dose of PBPs, no macroscopic or microscopic tumors were observed. We observed PBPs mediated dose-dependent decrease in oxidative DNA damage (8OHdG); inflammation (COX-2); proliferation (PCNA, Cyclin D1); expression of EGFR, and its downstream signaling kinases (pAkt, Akt, and mTOR); hypoxia (HIF1α) and angiogenesis (VEGF). There was also a PBPs mediated dose-dependent increase in apoptosis (Bax). Thus, our data clearly indicate that the observed chemopreventive potential of PBPs was due to modulation in the EGFR pathway associated with cell proliferation, hypoxia, and angiogenesis. Taken together, our results demonstrate preclinical chemopreventive efficacy of PBPs and give an insight into its mechanistic role in the chemoprevention of experimental oral cancer.

Prognostic Value of Radiological Extranodal Extension Detected by Computed Tomography for Predicting Outcomes in Patients With Locally Advanced Head and Neck Squamous Cell Cancer Treated With Radical Concurrent Chemoradiotherapy

Objective

Extra Nodal Extension (ENE) assessment in locally advanced head and neck cancers (LAHNCC) treated with concurrent chemo radiotherapy (CCRT) is challenging and hence the American Joint Committee on Cancer (AJCC) N staging. We hypothesized that radiology-based ENE (rENE) may directly impact outcomes in LAHNSCC treated with radical CCRT.

Materials and Methods

Open-label, investigator-initiated, randomized controlled trial (RCT) (2012–2018), which included LAHNSCC planned for CCRT. Patients were randomized 1:1 to radical radiotherapy (66–70 grays) with concurrent weekly cisplatin (30 mg/m²) [cisplatin radiation arm (CRT)] or same schedule of CRT with weekly nimotuzumab (200 mg) [nimotuzumab plus CRT (NCRT)]. A total of 536 patients were accrued and 182 were excluded due to the non-availability of Digital Imaging and Communications in Medicine (DICOM) computed tomography (CT) data. A total of 354 patients were analyzed for rENE. Metastatic nodes were evaluated based on five criteria and further classified as rENE as positive/negative based on three-criteria capsule irregularity with fat stranding, fat invasion, and muscle/vessel invasion. We evaluated the association of rENE and disease-free survival (DFS), loco-regional recurrence-free survival (LRRFS), and overall survival (OS).

Results

A total of 244 (68.9%) patients had radiologically metastatic nodes (rN), out of which 140 (57.3%) had rENE. Distribution of rENE was balanced in the two study groups CRT or NCRT (p-value 0.412). The median follow-up period was 39 months (ranging from 35.5 to 42.8 months). Complete response (CR) was seen in 204 (57.6%); incomplete response (IR), i.e., partial response plus stable disease (PR + SD), in 126 (35.6%); and progressive disease (PD) in 24 (6.8%). rENE-positive group had poor survival compared to rENE-negative group 3-year OS (46.7% vs. 63.6%), poor DFS (48.8% vs. 87%), and LRRFS (39.9% vs. 60.4%). rENE positive had 1.71 times increased risk of IR than rENE negative. Overall stage, site, clinical metastatic node (cN), response, and rENE were the significant factors for predicting OS, DFS, and LRRFS on univariate analysis. After making adjustment on multivariate analysis, rENE was an independent prognostic factor for DFS and trending to be significant for OS.

Conclusion

Pre-treatment rENE is an independent prognostic marker for survival in patients with LAHNSCC treated radically with CCRT that can be used as a potential predictive marker for response to treatment and hence stratify patients into responders vs. non-responders. We propose the mahajan rENE grading system applicable on CT, magnetic resonance imaging, positron emission tomography–contrast-enhanced CT, and ultrasound.

Abstract 6233: A deep learning pipeline to capture the prognostic immune responses in lymph nodes of breast cancer patients

Capturing tumor infiltrating leucocytes (TILS) and systemic immune responses in breast cancer informs disease progression and optimal treatment management. We have previously shown that morphological alterations in axillary lymph nodes (LNs), namely the formation of germinal centers in cancer-free LNs, adds prognostic value to TILs in triple negative breast cancer patients (TNBC) for the development of distant metastasis. Extending manual assessment of LNs beyond the detection of cancer requires the integration of robust deep learning pipelines into the digital pathology workflow. In this retrospective study, we used 1,100 Haematoxylin & Eosin-stained (H&E) Whole Slide Images (WSI) from Guy’s Hospital (London, UK) of metastatic and cancer-free LNs from 151 patients (100 N+) enriched for triple-negative or HER2-positive breast cancer to implement a supervised deep learning pipeline. A subset of 114 WSI, along with 5 breast cancer LN WSIs from each of Barts Hospital (London, UK) and Tianjin University Hospital (Tianjin, China), and 5 head and neck squamous cell carcinomas LN WSI (Guy’s Hospital) were used to develop, train and evaluate the segmentation task. For training Fully Convolutional Networks (FCNs), WSIs manually annotated for both germinal centers and sinuses formed a ground-truth set. Three FCNs were implemented: (i) a standard U-Net architecture; (ii) a U-Net model with an attention gate mechanism; and (iii) a multiscale-U-Net network (MSA-U-Net) that encodes, in parallel, a feature representation of the image at multiple resolutions. The MSA-U-Net achieved the best performance with an average dice score of 0.85 for germinal centers and 0.75 for sinuses. In comparison, the average dice score amongst 4 pathologists assessing 25 LN WSI for germinal centers and sinuses, was 0.67 and 0.61 respectively, demonstrating the robustness of the MSA-U-Net model. To quantify germinal centers and sinuses in LNs across the entire cohort, the trained MSA-U-Net was used in an inference step on all 1,100 WSI. The detected morphological features were initially localized within LNs using image thresholding and contouring techniques, and quantitatively assessed based on their number, area, shape, and Shannon diversity. We found significant morphological differences in metastatic and cancer-free LNs between N0 and N+ patients, with the latter displaying larger germinal centers with more irregular shapes especially in their metastatic LNs. In addition, we found differences in the Sinus area between LNs containing GCs and those without. Here, we propose a robust deep learning pipeline based on a multiscale FCN framework to automatically detect, localize and quantify histopathological immune features in WSI of LNs. By applying our pipeline to LNs of cancer patients, such as breast or head and neck, in prospective studies or clinical trials, we will further evaluate their prognostic and predictive values.

Citation Format: Gregory Verghese, Mengyuan Li, Amit Lohan, Nikhil Cherian, Swapnil Rane, Fangfang Liu, Aekta Shah, Pat Gazinska, Selvam Thavaraj, Amit Sethi, Anita Grigoriadis. A deep learning pipeline to capture the prognostic immune responses in lymph nodes of breast cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6233.

Author's Reply to "MoNuSAC2020: A Multi-Organ Nuclei Segmentation and Classification Challenge"

We had released MoNuSAC2020 as one of the largest publicly available, manually annotated, curated, multi-class, and multi-instance medical image segmentation datasets. Based on this dataset, we had organized a challenge at the International Symposium on Biomedical Imaging (ISBI) 2020. Along with the challenge participants, we had published an article summarizing the results and findings of the challenge (Verma et al., 2021). Foucart et al. (2022) in their "Analysis of the MoNuSAC 2020 challenge evaluation and results: metric implementation errors" have pointed ways in which the computation of the segmentation performance metric for the challenge can be corrected or improved. After a careful examination of their analysis, we have found a small bug in our code and an erroneous column-header swap in one of our result tables. Here, we present our response to their analysis, and issue an errata. After fixing the bug the challenge rankings remain largely unaffected. On the other hand, two of Foucart et al.'s other suggestions are good for future consideration, but it is not clear that those should be immediately implemented. We thank Foucart et al. for their detailed analysis to help us fix the two errors.

Prognostic Impact of Pattern of Mandibular Involvement in Gingivo-Buccal Complex Squamous Cell Carcinomas: Marrow and Mandibular Canal Staging System

Purpose

To study the pattern of mandibular involvement and its impact on oncologic outcomes in patients with gingivo-buccal complex squamous cell carcinoma (GBC-SCC) and propose a staging system based on the pattern of bone involvement (MMC: Marrow and mandibular canal staging system) and compare its performance with the 8th edition of the American Joint Committee on Cancer (AJCC8).

Methods

This retrospective observational study included treatment-naïve GBC-SCC patients who underwent preoperative computed tomography (CT) imaging between January 1, 2012, and March 31, 2016, at a tertiary care cancer center. Patients with T4b disease with high infratemporal fossa involvement, maxillary erosion, and follow-up of less than a year were excluded. The chi-square or Fisher’s exact test was used for descriptive analysis. Kaplan–Meier estimate and log-rank test were performed for survival analysis. Multivariate analysis was done using Cox regression analysis after making adjustments for other prognostic factors. p-Value <0.05 was considered as significant. Based upon the survival analysis with different patterns of bone invasion, a new staging system was proposed “MMC: Marrow and mandibular canal staging system”. “Akaike information criterion” (AIC) was used to study the relative fitted model of the various staging (TNM staging—AJCC8) with respect to survival parameters.

Results

A total of 1,200 patients were screened; 303 patients were included in the study. On radiology review, mandibular bone was involved in 62% of patients. The pattern of bone involvement was as follows: deep cortical bone erosion (DCBE) in 23%, marrow in 34%, and marrow with the mandibular canal in 43% of patients. Patients with DCBE and no bone involvement (including superficial cortical) had similar survival [disease-free survival (DFS) and locoregional recurrence-free survival (LRRFS)], and this was significantly better than those with marrow with or without mandibular canal involvement (for both DFS and LRRFS). Patients with DCBE were staged using the MMC, and when compared with the AJCC8, the MMC system was better for the prediction of survival outcomes, as AIC values were lower compared with those of the AJCC8. There was a significant association (p = 0.013) between the type of bone involvement and the pattern of recurrence.

Conclusions

For GBC-SCC, only marrow with or without mandibular canal involvement is associated with poorer survival outcomes. As compared with the AJCC8, the proposed Mahajan et al. MMC staging system downstages DCBE correlates better with survival outcomes.

Audit of Demographics, Treatment Patterns and Outcomes of Differentiated Thyroid Cancers Treated with Tyrosine Kinase Inhibitors

The real-world patterns of TKI use in differentiated thyroid cancer (DTC) are largely governed by the accessibility and financial feasibility of the patient with more sorafenib use compared to lenvatinib. There are limited data available on the toxicity profile, safety and tolerance of sorafenib and lenvatinib in DTC. Hence, we audited our practice on DTC. This is a retrospective single-centre analysis of patients with DTC who were referred to the Department of Medical Oncology for systemic therapy. Baseline demographics (age, sex, ECOG PS, comorbidities, substance use), tumour details (site of metastasis), previous treatment details, clinical features at metastasis (symptoms), the pattern of treatment, adverse events and outcomes including progression and death were extracted. There were 67 patients with DTC referred for systemic therapy; the median age was 56 (33-81) with a male preponderance (55.6%). The most common reason to start TKI therapy was radioactive iodine (RAI) cumulative dose > 600 milliCurie, followed by low iodine uptake in the RAI low-dose scan done at progression. The most common TKI used in the first line was sorafenib in 56 (83.6%) patients followed by lenvatinib in 9 (13.4%) patients. Papillary thyroid carcinoma was the most common histology (51, 76.1%), and the rest were follicular carcinoma (16, 23.9%). With a median follow-up of 36 months, the median PFS was 13.2 months (95% CI 10.4-16.0). The median OS was 18.8 months (95% CI 10.0-27.6). Among variables tested, no factors had a significant impact on the PFS or OS. The most common adverse events were hand-foot syndrome (54, 80.5%), diarrhoea (23, 33.3%) and transaminitis (24, 34.4%). The pattern of care of patients with RAI-refractory DTC is TKI therapy, especially sorafenib and lenvatinib in the real-world settings with comparable efficacy and safety profile compared to international literature.

An Audit of Systemic Therapy in Medullary Carcinoma Thyroid

There is a paucity of evidence of the impact of sorafenib on MCT and it is the preferred therapy used in India. We decided to do an audit of all patients of MCT who were referred to us for systemic therapy. The objective of this exercise was to identify the treatment pattern, outcomes, and adverse events with therapy in MCT. Baseline demographics (age, gender, ECOG PS, comorbidities, habits), tumor details (site of metastasis), previous treatment details, clinical features at metastasis (symptomatic or asymptomatic), the pattern of treatment, adverse events (CTCAE version 4.02), date of progression, date of death and status, and follow-up were extracted from the rare tumor database and electronic medical records. Out of 75 patients referred for therapy for MCT, 47 (62.7%) patients were considered for immediate tyrosine kinase inhibitors as they had symptomatic status and 28 (37.3%) patients were kept on observation due to the asymptomatic nature of the disease. Out of the 28 patients, 15 (53.6%, n = 28) patients were subsequently started on TKI while in 13 (46.4%, n = 28) patients observation was continued. In the overall cohort, the median PFS was 18.9 months (95% CI 11.9-29.9) and OS was 26.6 months (95% CI 14.4-39.0). Among variables tested, only female gender had an impact on PFS (hazard ratio = 0.364 95% CI 0.148-0.895; P = 0.028) and the absence of lung metastasis had a positive impact on OS (hazard ratio = 0.443 95% CI 0.207-0.95; P = 0.037). Most commonly used TKI was sorafenib (n = 61) and sunitinib in 1 patient. The most common adverse events with TKI were palmo-plantar dysesthesia (50, 80.6%) and oral mucositis (25, 40.2%). The strategy of treating symptomatic MCT and observing in asymptomatic MCT is associated with reasonable PFS and OS. Sorafenib is the most commonly used TKI in our setup and provides similar outcomes as globally.

Solitary Bone Plasmacytoma of Humerus Presenting as a Nonhealing Fracture in a Child: A Rare Entity

Solitary bone plasmacytoma is an extremely rare entity and is characterized by localized proliferation of monoclonal plasma cells. Plasmacytomas are extremely rare in the pediatric population. The median age at diagnosis is usually the fifth or sixth decade, with axial skeleton being more commonly involved than appendicular. We hereby, report the case of a 13-year-old boy with solitary bone plasmacytoma of the right humerus. Though extremely rare in the pediatric age group, plasmacytomas may be considered as one of the remote differentials in children presenting with solitary bone tumors.

MoNuSAC2020: A Multi-Organ Nuclei Segmentation and Classification Challenge

Detecting various types of cells in and around the tumor matrix holds a special significance in characterizing the tumor micro-environment for cancer prognostication and research. Automating the tasks of detecting, segmenting, and classifying nuclei can free up the pathologists' time for higher value tasks and reduce errors due to fatigue and subjectivity. To encourage the computer vision research community to develop and test algorithms for these tasks, we prepared a large and diverse dataset of nucleus boundary annotations and class labels. The dataset has over 46,000 nuclei from 37 hospitals, 71 patients, four organs, and four nucleus types. We also organized a challenge around this dataset as a satellite event at the International Symposium on Biomedical Imaging (ISBI) in April 2020. The challenge saw a wide participation from across the world, and the top methods were able to match inter-human concordance for the challenge metric. In this paper, we summarize the dataset and the key findings of the challenge, including the commonalities and differences between the methods developed by various participants. We have released the MoNuSAC2020 dataset to the public.

Robust Classification of Histology Images Exploiting Adversarial Auto Encoders

Deep learning (DL) thrives on the availability of a large number of high quality images with reliable labels. Due to the large size of whole slide images in digital pathology, patches of manageable size are often mined for use in DL models. These patches are variable in quality, weakly supervised, individually less informative, and noisily labelled. To improve classification accuracy even with these noisy inputs and labels in histopathology, we propose a novel method for robust feature generation using an adversarial autoencoder (AAE). We utilize the likelihood of the features in the latent space of AAE as a criterion to weigh the training samples. We propose different weighting schemes for our framework and evaluate the effectiveness of our methods on the publically available BreakHis and BACH histopathology datasets. We observe consistent improvement in AUC scores using our methods, and conclude that robust supervision strategies should be further explored for computational pathology.

Weakly supervised learning on unannotated hematoxylin and eosin stained slides predicts BRAF mutation in thyroid cancer with high accuracy

Deep neural networks (DNNs) that predict mutational status from H&E slides of cancers can enable inexpensive and timely precision oncology. Although expert knowledge is reliable for annotating regions informative of malignancy and other known histological patterns (strong supervision), it is unreliable for identifying regions informative of mutational status. This poses a serious impediment to obtaining higher prognostic accuracy and discovering new knowledge of pathobiology. We used a weakly supervised learning technique to train a DNN to predict BRAF V600E mutational status, determined using DNA testing, in H&E stained images of thyroid cancer tissue without regional annotations. Our discovery cohort was a tissue microarray of only 85 patients from a single hospital. Yet, on a large independent external cohort of 444 patients from other hospitals, the trained model gave an AUC = 0.98 (95% CI: 0.97-1.00), which is much higher than the previously reported results for detecting any mutation using H&E by DNNs trained using strong supervision. We also developed a visualization technique that can automatically highlight regions the DNN found most informative for predicting mutational status. Our visualization is spatially granular and highly specific in highlighting regions with strong negative and positive regions and move us towards explainable artificial intelligence. Using t-tests, we confirmed that the proportions of follicular or papillary histology and oncocytic cytology, as noted for each patient by a pathologist who was blinded to the mutational status, were significantly different between mutated and wildtype patients. However, based solely on these features noted by the pathologist, a logistic regression classifier gave an average AUC = 0.78 in 5-fold CV, which is much lower than that obtained using the DNN. These results highlight the potential of weakly supervised learning for training DNN models for problems where the informative visual patterns and their locations are not known a priori. This article is protected by copyright. All rights reserved.

Carcinoma Cuniculatum of the Oral Cavity: A Series of 6 Cases and Review of Literature

Carcinoma cuniculatum (CC) is a rare variant of squamous cell carcinoma (SCC) that is characterized by minimal cytologic atypia and a unique deeply infiltrative growth pattern resembling rabbit burrows (cuniculi). With less than 75 cases reported in the head and neck, the clinical and pathologic spectrum of this entity remains poorly understood. A retrospective review of the clinical and pathologic features of archival cases of oral CC was performed. A total of six cases of oral CC were identified. Age ranged from 25-77 years; the male-to-female ratio was 5:1. All patients had a long-standing history of tobacco and betel-quid consumption. The tumors were distributed in the gingivobuccal sulcus (n = 2), the tongue (n = 2), buccal mucosa (n = 1), and the palate (n = 1). Histology in all cases typically revealed a tumor composed of well-differentiated squamous epithelium, devoid of atypia, lining deeply infiltrative, large-sized, branching, keratin-filled cavities, resembling rabbit-burrows. Dense lymphocytic infiltrates and discharging micro-abscesses were regular features. Underlying bone invasion and lymph node metastasis were observed in 1 patient. One patient with a tongue tumor developed locoregional recurrence at 10 months while none developed distant metastasis. Oral CC is a rare and under-recognized variant of SCC with locally aggressive behavior. Lack of familiarity with this variant exacerbated by the absence of cytologic anaplasia makes CC susceptible to multiple negative biopsies and erroneous diagnoses. Awareness of this clinicopathologic entity is essential to allow its accurate diagnosis and optimal management.

Multifocal Pseudomyogenic Hemangioendothelioma Involving the Scalp and Nose, Misdiagnosed as A Sarcoma: A Rare Case Report

This case report aims to present clinicopathological features of an extremely rare case of multifocal pseudomyogenic hemangioendothelioma (PMHE) in the scalp. A 21-year-old male developed multiple, focally ulcerated, nodules over the root of his nose and scalp. One of the skin lesions was sampled at another dermatology clinic, where this was diagnosed as a sarcoma. A review of biopsy sections showed well-circumscribed dermal lesions, comprising plump spindle and epithelioid cells, mimicking rhabdomyoblasts. Immunohistochemically, tumor cells were positive for AE1/AE3, CD31, FLI-1 and ERG. INI-1 was retained. A diagnosis of PMHE was offered. Subsequently, the patient underwent wide excision and has been asymptomatic for 8 months, post-surgery. PMHE is rarely reported in the head and neck region, where it can constitute a diagnostic pitfall. Awareness of this tumor and appropriate immunohistochemical stains are necessary for its timely diagnosis, in order to avoid radical treatments. A review of similar, previously documented cases is presented.

Intensity-modulated radiation therapy for nasal cavity and paranasal sinus tumors: Experience from a single institute

BACKGROUND

To assess the efficacy of intensity-modulated radiation therapy (IMRT) for tumors of the nasal cavity and paranasal sinus (PNS) region.

MATERIALS AND METHODS

Two hundred fourteen patients with tumors of the nasal cavity and PNS region treated with curative intent IMRT between 2007 and 2019 were included in this retrospective analysis.

RESULTS

Fifty-one (24.1%) received definitive RT/CTRT and 163 (75.9%) received adjuvant RT. Most common histology was squamous cell carcinoma (26.1%) followed by adenoid cystic carcinoma (21.5%). The median follow-up was 43.5 months. The 5-year local control (LC), event-free survival (EFS), and overall survival (OS) for the entire cohort was 66.9%, 59%, and 73.9%, respectively. On univariate analysis treatment with nonsurgical modality, T classification and undifferentiated/poorly differentiated histology were associated with inferior 5-year LC, EFS, and OS. Four patients had late Grade 3/Grade 4 ocular toxicity.

CONCLUSIONS

IMRT should be the standard of care for tumors of PNS region across all histologies and treatment setting.

Abstract PO-014: Deep learning-based segmentation accurately captures histological features in cancer-free lymph nodes of breast cancer patients

Background Histological changes in the cancer-free lymph nodes (LNs) can add to the risk prediction of developing distant metastasis for LN-positive breast cancer patients. We have previously shown that the formation of germinal centres and enlarged sinus are premetastatic features and indicative of superior prognosis. As Whole Slide Images (WSI) demonstrated suitability for detection of LN metastasis with high accuracy based on Convolutional Neural Networks (CNN), we extended the CNN assessments to the detection and segmentation of these premetastatic features. Materials For 136 breast cancer patients, retrospectively collected from Guy’s Hospital, 1,422 Whole Slide Images (WSI) of axillary LNs (containing 1 to 3 LNs per WSI) and primary invasive tumours were 40x scanned by the Hamamatsu Nanozoomer. Two pathologists manually annotated germinal centres and sinuses in 111 of these images facilitating the ground-truth masks for a binary segmentation task using deep learning models. Results Each WSI was split into a series of overlapping patches along with corresponding ground-truth masks at 2.5x, 5x and 10x. Fully Convolutional Neural Networks (FCNs) based on the U-Net architecture were trained on patches from 95 annotated WSI to perform a segmentation for both germinal centres and sinus, with the remaining images used for validation (n= 6) and test sets (n= 10). In addition to a vanilla U-Net encoder-decoder architecture, a model with an attention mechanism to filter feature maps along skip connections was also explored. Additionally, a U-Net with dilated convolutions at each convolution block was used in a multi-scale mechanism that combines features from three different resolutions. To increase the robustness of the FCNs, training data was augmented using random image rotation and different colour perturbation transformations. Predictions were made on regions of WSIs containing whole LNs and were compared with the pathologist annotations on holdout testing images. The Dice coefficient was used to compare the predicted segmentations with the ground-truth annotations. Standard U-Net and attention U-Net architectures achieved an average Dice coefficient of 0.75 across the holdout testing images at 2.5x magnification for germinal centres, with individual images achieving a Dice score above 0.8. Multiscale models achieved the best results at 10x magnification with an average Dice score of 0.85 for germinal centres and 0.75 for sinus. The trained models were then applied to &gt; 1,300 unseen and non-annotated WSIs of axillary LNs. Conclusion Our developed FCN models allow an automated, precise, reproducible and time-efficient approach to annotate germinal centres and sinuses in cancer-free and metastatic LNs. Trained models are currently applied to &gt;1,300 unseen and non-annotated WSIs of axillary LNs, and will demonstrate their comparable performance with manual pathological assessment.

Citation Format: Gregory Verghese, Anita Grigoriadis, Amit Sethi, Amit Lohan, Nikhil Cherian, Swati Meena, Harry Chinque, Fangfang Liu, Ed Martin, Tristan Clark, Cheryl Gillet, Julie Owen, Swapnil Rane. Deep learning-based segmentation accurately captures histological features in cancer-free lymph nodes of breast cancer patients [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-014.

Remote Reporting from Home for Primary Diagnosis in Surgical Pathology: A Tertiary Oncology Center Experience during the COVID-19 Pandemic

BACKGROUND

The COVID-19 pandemic accelerated the widespread adoption of digital pathology (DP) for primary diagnosis in surgical pathology. This paradigm shift is likely to influence how we function routinely in the postpandemic era. We present learnings from early adoption of DP for a live digital sign-out from home in a risk-mitigated environment.

MATERIALS AND METHODS

We aimed to validate DP for remote reporting from home in a real-time environment and evaluate the parameters influencing the efficiency of a digital workflow. Eighteen pathologists prospectively validated DP for remote use on 567 biopsy cases including 616 individual parts from 7 subspecialties over a duration from March 21, 2020, to June 30, 2020. The slides were digitized using Roche Ventana DP200 whole-slide scanner and reported from respective homes in a risk-mitigated environment.

RESULTS

Following re-review of glass slides, there was no major discordance and 1.2% (n = 7/567) minor discordance. The deferral rate was 4.5%. All pathologists reported from their respective homes from laptops with an average network speed of 20 megabits per second.

CONCLUSION

We successfully validated and adopted a digital workflow for remote reporting with available resources and were able to provide our patients, an undisrupted access to subspecialty expertise during these unprecedented times.

FISH patterns of ROS1, MET, and ALK with a correlation of ALK immunohistochemistry in lung cancer: a case for introducing ALK immunohistochemistry 'Equivocal' interpretation category in the Ventana anti-ALK (D5F3) CDx assay - A tertiary cancer center experience

Background

Mutations in ROS1, ALK, and MET genes are targetable alterations in non-small cell lung cancer (NSCLC). They can be evaluated by different techniques, most commonly fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC).

Methods

We explored the prevalence of ROS1, ALK, MET mutations, discuss clinicopathological associations and FISH signal patterns on 413 consecutive cases of EGFR negative lung carcinoma from March 2016 to April 2017 using FISH for ALK, ROS1, and MET along with ALK (D5F3) IHC.

Results

ROS1 gene rearrangement, ALK positivity (IHC and/or FISH), and MET amplification were seen in 18/358 (5%) cases, 76/392 cases (19.4%), and 10/370 (2.7%) cases, respectively. ALK FISH and ALK IHC were positive in 51/300 (17%) and 58/330 cases (17.57%), respectively, while 8/330 (2.4%) cases were ALK IHC "equivocal" of which 3/8 (37.5%) were ALK FISH positive. Of ALK FISH and IHC co-tested cases, 43/238 (18.07%) cases were positive by both techniques, while 15/43 (34.88%) of ALK positive cases showed discordant ALK FISH and IHC results. All ROS1 rearranged and MET amplified cases were adenocarcinoma. Signet ring cell histology was associated with 78.57% likelihood of being either ALK or ROS1 positive. Genomic heterogeneity was seen in 30% of MET amplified cases.

Conclusions

ALK/ROS1/MET gene alterations were found in 25.18% of NSCLC cases. An ALK IHC "equivocal" interpretation category should be incorporated into practice. Atypical patterns of ROS1 and genomic heterogeneity need to be evaluated further for any clinical relevance.

Parathyroid Carcinoma: A Single-Institution Experience with an Emphasis on Histopathological Features

Parathyroid carcinoma (PC) is a rare malignancy that poses a diagnostic challenge on histologic examination. We analyzed various clinicopathologic features of PC. Pathology reports and slides were reviewed to evaluate the diagnostic histopathologic features of archived cases of PC from the years of 2004-2018. The study cohort comprised twenty cases of PC. The median age was 49 years (range 21-73 years) with equal gender distribution (M:F = 1:1). Most patients presented with symptoms of hypercalcemia (n = 7, 54%). Serum calcium and serum parathyroid hormone were elevated in all but one patient. The right inferior parathyroid was commonly involved (n = 8/14, 57%). The mean tumor size was 2.4 cm (range 0.8-3.5 cm). On frozen section examination, PC was diagnosed in 8 out of 9 cases. Vascular (n = 19/20, 95%) and soft tissue invasion (n = 10/20, 50%) were the most common characteristic histologic findings. Capsular invasion was identified in all cases. Perineural invasion or metastasis at presentation was absent in all cases. Other histological features noted were intratumoral fibrous bands (70%), nodular growth pattern (70%), moderate nuclear atypia (30%), prominent nucleoli (20%), and necrosis (20%). Regional lymph nodes were negative for metastatic disease in all cases (n = 10). Eight out of 16 patients received adjuvant radiotherapy. Follow-up was available in 16 cases (median 21.5 months). Two patients died of disease. Vascular and soft tissue invasion are the most common diagnostic histologic features of PC. Capsular invasion is important to distinguish PC from its benign counterparts. Intraoperative frozen section examination can be used for accurate diagnosis and surgical management.

Prognostic and predictive significance of nuclear HIF1α expression in locally advanced HNSCC patients treated with chemoradiation with or without nimotuzumab

BACKGROUND

Anti-EGFR-based therapies have limited success in HNSCC patients. Predictive biomarkers are greatly needed to identify the patients likely to be benefited from these targeted therapies. Here, we present the prognostic and predictive association of biomarkers in HPV-negative locally advanced (LA) HNSCC patients.

METHODS

Treatment-naive tumour tissue samples of 404 patients, a subset of randomised Phase 3 trial comparing cisplatin radiation (CRT) versus nimotuzumab plus cisplatin radiation (NCRT) were analysed to evaluate the expression of HIF1α, EGFR and pEGFR by immunohistochemistry and EGFR gene copy change by FISH. Progression-free survival (PFS), locoregional control (LRC) and overall survival (OS) were estimated by Kaplan-Meier method. Hazard ratios were estimated by Cox proportional hazard models.

RESULTS

Baseline characteristics of the patients were balanced between two treatment groups (CRT vs NCRT) and were representative of the trial cohort. The median follow-up was of 39.13 months. Low HIF1α was associated with better PFS [HR (95% CI) = 0.62 (0.42-0.93)], LRC [HR (95% CI) = 0.56 (0.37-0.86)] and OS [HR (95% CI) = 0.63 (0.43-0.93)] in the CRT group. Multivariable analysis revealed HIF1α as an independent negative prognostic biomarker. For patients with high HIF1α, NCRT significantly improved the outcomes [PFS:HR (95% CI) = 0.55 (0.37-0.82), LRC:HR (95% CI) = 0.55 (0.36-0.85) and OS:HR (95% CI) = 0.54 (0.36-0.81)] compared to CRT. While in patients with low HIF1α, no difference in the clinical outcomes was observed between treatments. Interaction test suggested a predictive value of HIF1α for OS (P = 0.008).

CONCLUSIONS

High HIF1α expression is a predictor of poor clinical response to CRT in HPV-negative LA-HNSCC patients. These patients with high HIF1α significantly benefited with the addition of nimotuzumab to CRT.

CLINICAL TRIAL REGISTRATION

Registered with the Clinical Trial Registry of India (Trial registration identifier-CTRI/2014/09/004980).

Two Cases of Intraosseous Pseudomyogenic (Epithelioid Sarcoma-Like) Hemangioendothelioma With Unusual Features, Expanding the Clinicopathological Spectrum

Pseudomyogenic (epithelioid sarcoma-like) hemangioendothelioma (PMHE) is a distinct vascular neoplasm mostly observed in the lower extremities of young adults with characteristic histopathological features. In this article, we present 2 unusual cases of PMHE. Case 1: A 28-year-old male presented with pain and stiffness in his left shoulder. Radiologic examination revealed an expansile, lytic, heterogeneously enhancing, destructive lesion in his left scapula, along with multiple avid marrow lesions in his other bones. Biopsy revealed a cellular tumor composed of plump spindly and epithelioid cells arranged in fascicles and focally, in clusters, admixed with neutrophils and histiocytes, with interspersed many osteoclast-like giant cells and reactive bone. Case 2: A 63-year-old male presented pain and swelling in his right middle finger, with no other lesion elsewhere in his body. Radiologic imaging disclosed a 1.5-cm-sized lobulated, expansile, lytic, destructive lesion in the distal end of the third metacarpal bone of his right hand. Biopsy examination revealed a cellular spindle cell tumor, composed of plump spindly cells, arranged in fascicles, including "tadpole-like" or "strap-like" cells and interspersed inflammatory cells. Immunohistochemically, tumor cells in both cases were positive for AE1/AE3, CD31, and Fli1, while negative for desmin and CD34. INI11 was retained. The presented cases of PMHE, occurring at unusual locations, in an older individual in the second case, along with variable histopathological features, noted in the first case, seem to expand the clinicopathological spectrum of these uncommon neoplasms. Differential diagnoses and review of similar cases are presented.

Multimodality imaging manifestations of Rosai-Dorfman disease

Background

Rosai-Dorfman disease (RDD) is a rare lympho-histiocytic disorder of indeterminate etiology usually presenting with lymph node involvement, and infrequently with extra-nodal manifestations. The diagnosis of this condition is challenging due to the wide spectrum of disease manifestations.

Purpose

To elucidate the radiologic features of this disease using multimodality imaging in histopathologically proven cases and to identify characteristic features that would enable its differentiation from its mimics.

Material and Methods

We retrospectively evaluated imaging studies of 19 patients with histopathologically confirmed RDD presenting to our institute between January 2004 and March 2016. Imaging modalities included magnetic resonance imaging (MRI), computed tomography, FDG-positron emission tomography (PET) CT, mammography, and ultrasonography.

Results

Lymphadenopathy was the most common imaging feature in our study, seen in 11 (57.8%) cases followed by sino-nasal involvement in 7 (36.8%) cases and intracranial masses in 5 (26.3%) cases. Bilateral homogeneously enhancing cervical lymphadenopathy with avidity on FDG-PET scans was the predominant abnormality on imaging. Sino-nasal involvement manifested as homogeneously enhancing soft-tissue masses occupying the paranasal sinuses. Intracranial disease manifested as sellar/suprasellar masses, dural-based lesions along the cerebral hemispheres and choroid plexus enlargement. Unusual disease manifestations included spinal, osseous, and breast lesions.

Conclusion

Due to the high likelihood of multifocal involvement, the recognition of RDD at one site necessitates screening of other sites for disease. Homogeneously enhancing, FDG-avid lymphadenopathy and sino-nasal masses in association with hypointense extra-nodal lesions on T2-weighted MRI are imaging features which could aid the diagnosis of RDD and facilitate its differentiation from pathologies that present in a similar manner.

Deep Learning to Estimate Human Epidermal Growth Factor Receptor 2 Status from Hematoxylin and Eosin-Stained Breast Tissue Images

Context:

Several therapeutically important mutations in cancers are economically detected using immunohistochemistry (IHC), which highlights the overexpression of specific antigens associated with the mutation. However, IHC panels can be imprecise and relatively expensive in low-income settings. On the other hand, although hematoxylin and eosin (H&E) staining used to visualize the general tissue morphology is a routine and low cost, it does not highlight any specific antigen or mutation.

Aims:

Using the human epidermal growth factor receptor 2 (HER2) mutation in breast cancer as an example, we strengthen the case for cost-effective detection and screening of overexpression of HER2 protein in H&E-stained tissue.

Settings and Design:

We use computational methods that reliably detect subtle morphological changes associated with the over-expression of mutation-specific proteins directly from H&E images.

Subjects and Methods:

We trained a classification pipeline to determine HER2 overexpression status of H&E stained whole slide images. Our training dataset was derived from a single hospital containing 26 (11 HER2+ and 15 HER2–) cases. We tested the classification pipeline on 26 (8 HER2+ and 18 HER2–) held-out cases from the same hospital and 45 independent cases (23 HER2+ and 22 HER2–) from the TCGA-BRCA cohort. The pipeline was composed of a stain separation module and three deep neural network modules in tandem for robustness and interpretability.

Statistical Analysis Used:

We evaluate our trained model through area under the curve (AUC)-receiver operating characteristic.

Results:

Our pipeline achieved an AUC of 0.82 (confidence interval [CI]: 0.65–0.98) on held-out cases and an AUC of 0.76 (CI: 0.61–0.89) on the independent dataset from TCGA. We also demonstrate the region-level correspondence of HER2 overexpression between a patient's IHC and H&E serial sections.

Conclusions:

Our work strengthens the case for automatically quantifying the overexpression of mutation-specific proteins in H&E-stained digital pathology, and it highlights the importance of multi-stage machine learning pipelines for added robustness and interpretability.

Prognostic value of radiological extranodal extension detected by computed tomography for predicting outcomes in head and neck squamous cell cancer patients treated with radical chemoradiotherapy

6560

Background: As per the AJCC 8th edition ENE/ECS is the most important predictor for N staging of HNSCC and is one of the key predictor of outcomes. Because ENE/ECS is based on pathological findings after surgery and it is difficult to predict outcomes for locally advanced squamous head and neck cancer (LASHNC) treated radically with CCRT. We hypothesized that ENE assessed by CT imaging (rENE) may directly correlate with outcomes in LASHNC treated radically with CCRT. Methods: This open-label, investigator-initiated, phase 3, randomized trial was conducted from 2012 to 2018. Adult patients with LASHNC who were fit for radical chemoradiation were randomized 1:1 to receive either radical radiotherapy (66-70 grays) with concurrent weekly cisplatin (30 mg/m2) (CRT) or the same schedule of CRT with weekly nimotuzumab (200 mg) (NCRT). 536 patients were accrued,182 were excluded due to non-availability DICOM CT scan, 354 patients were analysed for rENE (based on 6 criterion for metastasis and 3 for rENE). Near equal distribution of patients was achieved in CRT arm (170 patients) and NCRT arm (184 patients). There were 181 (51.1%) oropharynx and 173(48.9%) larynx and hypopharynx patients. We evaluated association of radiological ENE and clinical outcomes.The endpoints were disease-free survival (DFS), duration of locoregional control (LRC), and overall survival (OS). Results: There were 244(68.9%) patients with radiologically metastatic nodes, out of which 140(57.3%) had rENE. There was no significant association between rENE and CRT (p value 0.3) or NCRT (p value 0.412). The median follow-up was 33.0 months (95%CI 30.7-35.2 months). Complete response was achieved in 204 (57.6%) cases, PR/SD in 126(35.6%) cases and PD in 24(6.8%) cases. rENE positive patients had poor overall 3-year survival (46.7%), poor DFS (48.8%) and LRC (39.9%) than rENE negative cases (63.6%, 87%, 60.4%). rENE positive cases had 1.71 times increase chances of incomplete response than rENE negative cases. Overall stage, clinical positive node, response, rENE and site were the only significant factors for predicting OS, DFS and LRC. Conclusions: In conclusion, pre-treatment rENE can be regarded as an independent prognostic factor for survival (OS, DFS, LRC) in patients with LASHNC treated radically with CCRT. Pre-treatment rENE is not only associated with CCRT response but is also associated with poor prognosis and hence rENE, as an imaging biomarker, can stratify responder’s vs non-responders. Clinical trial information: CTRI/2014/09/004980 .

Deep learning-based predictive imaging biomarker model for EGFR mutation status in non-small cell lung cancer from CT imaging

3106

Background: Deep learning based radiogenomic (DLR) models present a promising performance in assisting lung cancer care. The purpose of this study was 1) To develop and validate DLR signatures to predict the EGFR mutation, 2) To assess the incremental value of these DLR signatures in comparison to the traditional clinical and semantic features. Methods: 223 patients were selected from two phase III randomized trials in patients with advanced non-squamous NSCLC with EGFR-sensitizing mutation and EGFR wild type who were planned to receive palliative therapy (trial 1: gefitinib or gefitinib plus pemetrexed and carboplatin and trial 2: pemetrexed maintenance and erlotinib maintenance). Our method is an end-to-end pipeline that requires only the manually selected tumour region in a CT image without precise tumour boundary segmentation or human-defined features. Two deep convolutional neural networks with 3D U-Net architectures are trained to segment lung masses and nodules from 3D regions of the CT image. The primary end point was EGFR prediction using Radiomics and DLR pipeline. We also compared the performance of combination of models in predicting the mutational status. Results: A total of 223 patients (mean age, 54.18 years; age range, 28–80 years) were included in this study. There were 121 (54.3%) patients with EGFR mutation and 102 (45.7%) patients who were EGFR wild type. On multivariate logistic regression analysis, Clinical variable and CT semantic features that were found to be significantly associated EGFR mutation were tumor stage, smoking status, pure solid texture, presence of non-tumor lobe nodule, and average enhancement. For predicting EGFR mutation, ROC curve plotted with clinical variables model, CT semantic variables model, Radiomics model, DLR model showed an AUC value of 0.70, 0.73, 0.94, 0.72 respectively. Clinical variables and semantic features were added to the radiomics predictive model and deep learning predictive model independently, showed further improvement in the accuracy for either model from AUC 0.94+/-0.02 to 0.96+/-0.02 and from AUC 0.72+/-0.02 to 0.82+/-0.04 respectively. Conclusions: The radiomics and DLR model by machine-learned information, extracted from CT images without precise manual segmentation, could predict EGFR mutation with very high accuracy. This AI based model can be used as non-invasive and easy-to-use surrogate imaging biomarker for EGFR mutation status prediction. Clinical trial information: CTRI/2018/10/022102 .

A Novel Approach for Fully Automatic Intra-Tumor Segmentation With 3D U-Net Architecture for Gliomas

Purpose: Gliomas are the most common primary brain malignancies, with varying degrees of aggressiveness and prognosis. Understanding of tumor biology and intra-tumor heterogeneity is necessary for planning personalized therapy and predicting response to therapy. Accurate tumoral and intra-tumoral segmentation on MRI is the first step toward understanding the tumor biology through computational methods. The purpose of this study was to design a segmentation algorithm and evaluate its performance on pre-treatment brain MRIs obtained from patients with gliomas.

Materials and Methods: In this study, we have designed a novel 3D U-Net architecture that segments various radiologically identifiable sub-regions like edema, enhancing tumor, and necrosis. Weighted patch extraction scheme from the tumor border regions is proposed to address the problem of class imbalance between tumor and non-tumorous patches. The architecture consists of a contracting path to capture context and the symmetric expanding path that enables precise localization. The Deep Convolutional Neural Network (DCNN) based architecture is trained on 285 patients, validated on 66 patients and tested on 191 patients with Glioma from Brain Tumor Segmentation (BraTS) 2018 challenge dataset. Three dimensional patches are extracted from multi-channel BraTS training dataset to train 3D U-Net architecture. The efficacy of the proposed approach is also tested on an independent dataset of 40 patients with High Grade Glioma from our tertiary cancer center. Segmentation results are assessed in terms of Dice Score, Sensitivity, Specificity, and Hausdorff 95 distance (ITCN intra-tumoral classification network).

Result: Our proposed architecture achieved Dice scores of 0.88, 0.83, and 0.75 for the whole tumor, tumor core and enhancing tumor, respectively, on BraTS validation dataset and 0.85, 0.77, 0.67 on test dataset. The results were similar on the independent patients' dataset from our hospital, achieving Dice scores of 0.92, 0.90, and 0.81 for the whole tumor, tumor core and enhancing tumor, respectively.

Conclusion: The results of this study show the potential of patch-based 3D U-Net for the accurate intra-tumor segmentation. From experiments, it is observed that the weighted patch-based segmentation approach gives comparable performance with the pixel-based approach when there is a thin boundary between tumor subparts.

Soft tissue deposit in neck dissection specimen carries a poor prognosis in oral cancer: A matched pair analysis

BACKGROUND

Soft tissue deposit (STD) is the presence of tumor foci in the neck at sites other than lymphoid tissue. We evaluated the factors associated with STD and their impact on outcomes.

METHODS

The records of 4812 treatment naïve oral cancers operated between December 2010 and September 2015 were screened for the presence of STD. A matched-pair analysis was performed to obtain 450 patients without STD. Chi-square was used to test association with clinicopathological factors and hazard ratio (HR) for overall survival (OS) and disease-free survival (DFS) were calculated.

RESULTS

STD was present in 0.6% and associated with gingiva-buccal cancers, advanced stage, poor differentiation, presence of extranodal extension (ENE) and higher locoregional/distant metastasis. They had a significantly poorer OS (HR-2.08; P-value .0073) and DFS (HR-2.5; P-value .0002) with and without ENE [HR-1.96 and 3.7, respectively].

CONCLUSION

Despite aggressive adjuvant therapy STD has a significant negative impact on outcomes.

Serendipitous actinomycosis infection mimicking malignant lesion in carcinoma of unknown primary

Actinomycosis is a bacterial infection caused by Actinomyces species that are commensal in the human oral cavity, digestive tract, and genital tract. The present case highlights actinomycosis of the hypopharynx, mimicking malignancy in ¹⁸F-fluorodeoxyglucose positron emission tomography-computed tomography.

Oral administration of 3,3'-diselenodipropionic acid prevents thoracic radiation induced pneumonitis in mice by suppressing NF-kB/IL-17/G-CSF/neutrophil axis

The incidence of symptomatic radiation induced lung pneumonitis (RILP), a major dose limiting side effect of thoracic radiotherapy, is in the range of 15-40%. Therapeutic options for the prevention and treatment of RILP are limited. Hence there is a need for developing novel radioprotectors to prevent RILP which can be patient compliant. This study sought to evaluate the efficacy of oral 3,3'-diselenodipropionic acid (DSePA), a novel selenocystine derivative to prevent RILP. C3H/HeJ (pneumonitis responding) mice received a single dose of 18 Gy, whole thorax irradiation and a subset were treated with DSePA orally (2.5 mg/kg), three times per week beginning 2 h post irradiation and continued till 6 months. DSePA delayed onset of grade ≥ 2 RILP by 45 days compared to radiation control (~105 versus ~60 days). It also reversed the severity of pneumonitis in 3/10 radiation treated mice leading to significant improvement in asymptomatic survival compared to radiation control (~180 versus ~102 days). DSePA significantly (p < 0.05) reduced the radiation-mediated infiltration of polymorphonuclear neutrophils (PMN) and elevation in levels of cytokines such as IL1-β, ICAM-1, E-selectin, IL-17 and TGF-β in the bronchoalveolar lavage fluid. Moreover DSePA lowered PMN-induced oxidants, maintained glutathione peroxidase activity and suppressed NF-kB/IL-17/G-CSF/neutrophil axis in the lung of irradiated mice. Additionally, this compound did not protect A549 (lung cancer) derived xenograft tumor from radiation exposure in SCID mice. DSePA offers protection to normal lung against RILP without affecting radiation sensitivity of tumors. It has the potential to be developed as an oral agent for preventing RILP.

P6378Characteristics of hypertensive iron ore miners with oral glucocorticoid induced diabetes mellitus and compromised lung function

Introduction

Oral glucocorticoids are efficacious agents for treating respiratory diseases, with high risk to induce metabolic diseases including diabetes and hypertension

Purpose

To analyse the hypertensive patient population from the rural setting on ongoing glucocorticoid treatment and diagnosed as diabetic due to the extensive chronic methylprednisolone equivalent glucocorticoid dose of more than 4 mg/day for at least 3 months for treatment of respiratory ailments due to occupational hazard

Methods

We diagnosed the identified and analysed a subset of patients from the Redi iron ore miners (n=25,500), who were newly diagnosed diabetics (n=1100) and with associated systemic hypertension (systolic BP >140 mmHg, diastolic BP > than 90 mmHg). Patients diagnosed as Oral Glucocorticoid Induced Diabetes Mellitus (GIDM) and hypertensives (n=847; 572 males and 275 females) were characterised on metabolic parameters. t-test was utilised for statistical analysis

Results

We observed a high incidence of GIDM (4.31%) with a high rate of hypertension (77%, 847/1100). Mean SBP was 150 mmHg (SD ± 9.9, 95% CI 147–152), mean DBP was 90 mmHg (SD ± 8.9, 95% CI 88–92). BMI in GIDM with hypertension (GIDM-H) group was similar to GIDM group (26 kg/m2) (Table), with relatively older age in GIDM–H (53 vs 51 years). The glycemic parameters were relatively more deranged in GIDM-H group. There was significant correlation for a high propensity of relatively younger <50 years (n=506) and near normal weight people with BMI <26 kg/m2 (n=594) to develop GIDM (p=0.006)

Metabolic Parameters GIDM Vs GIDM-H Mean ±SD, min, max, 95% CI GIDM (n=1100) GIDM with Hypertension (n=847) Age (yrs) 51 (SD ±14, 95% CI 48–54) 53 (SD ±13, 95% CI 50–56) HbA1c (%) 7.1 (SD ±0.62, 95% CI 7–7.3), 7.2 (SD ±0.60, 95% CI 7.1–7.3) BMI (kg/m2) 26 (SD ±3.6, 95% CI 25–27) 26 (SD ±3.4, 95% CI 25–27) Fasting Plasma Glucose (mg/dl) 146 (SD ±49, 95% CI 136–155) 149 (SD ±48, 95% CI 138–160) Post Prandial Glucose (mg/dl) 207 (SD ±74, 95% CI 192 - 221) 212 (SD ±74, 95% CI 195–228)

Conclusions

The high rates of hypertension in patients with GIDM is an alarmingly problem which has pronounced harmful comorbid implications, especially in younger, economically productive population in limited resource setting. The health care professionals need to be educated to limit use of glucocorticoids to inhalational therapy with minimal systemic effects. Specific pathophysiological approach to address the steroid induced insulin resistance, strategies to reduce cardiovascular risk and damage, makes treatment of hypertension and GIDM difficult, which calls for action to limit the mining activities and mitigate risk of chronic complications

Correction to: Tumour origin and R1 rates in pancreatic resections: towards consilience in pathology reporting

The authors regret that one of the author's given name was missing and a typographical error was present in Reference 26 of the above article. These are presented correctly in this article.

Tumour origin and R1 rates in pancreatic resections: towards consilience in pathology reporting

To evaluate differences in the R1 rates of ampullary (AC), pancreatic (PC), and distal bile duct (DBD) cancers in pancreatoduodenectomies (PD) using standardised pathology assessment. Data of PD (2010-2011) analysed in accordance with the Royal College of Pathologists (UK) protocol, were retrieved. Clinicopathologic features, including frequency, topography, and mode of margin involvement in AC (n = 87), PC (n = 18), and DBD (n = 5) cancers were evaluated. The R1 rate was 7%, 67%, and 20% in the AC, PC, and DBD cancers (p < 0.001). Within the PC cohort, R1 rate was heterogeneous (chemo-naïve, 77%; post-neoadjuvant, 40%). Commonest involved margins were as follows: posterior in overall PD (35%), AC (43%), overall PC (33%), and post-neoadjuvant PC (100%); superior mesenteric artery margin in chemo-naïve PC (38%) and common bile duct margin in DBD (100%) cancers. In AC, majority (66%) of R1 were signet ring cell type. Indirect margin involvement due to tumour within lymph node, perineural sheath or lymphovascular space was observed in 26% cases, and altered R1 rate in AC, PC, and DBD cohorts by 1%, 12%, and 0%, respectively. Although not statistically significant, patients with R1 had lower disease-free survival than those with R0 (mean, 25.4 months versus 44.4 months). Tumour origin impacts R1 data in PD necessitating its accurate classification by pathologists. Indirect involvement, histology, and neoadjuvant therapy influence the R1 rate, albeit in a minority of cases. Generating cogent R1 data based on standardised pathology reporting is the foremost need of the hour.

Poorly Differentiated Chordomas Showing Loss of INI1/SMARCB1: A Report of 2 Rare Cases With Diagnostic Implications

Poorly differentiated chordomas are rare musculoskeletal tumors. Case 1. A 42-year-old lady presented with quadriparesis of 2 months' duration. Radiologic imaging disclosed a soft tissue mass in her left prevertebral- and paravertebral cervical region. Case 2. A 4-year-old male child presented with neck pain and restricted head movements of 1-year duration. Radiologic imaging revealed a contrast enhancing, paraspinal soft tissue mass in his cervical region. Microscopic examination in both the cases revealed a cellular malignant tumor composed of moderate to markedly pleomorphic cells with interspersed mitotic figures, along with focal myxoid change and necrosis. By immunohistochemistry, tumor cells in both cases were diffusely positive for pan cytokeratin (AE1/AE3) and brachyury, whereas these were negative for INI1/SMARCB1. Tumor cells in the second case were also positive for glypican3. The first case developed pulmonary metastasis, while the second case developed recurrence. Poorly differentiated chordomas are uncommon tumors, invariably characterized by loss of INI1. These tumors can be rarely seen in adults and need to be differentiated from their diagnostic mimics, in view of treatment implications and their relatively aggressive clinical outcomes.