Weakly Supervised Learning using Attention gates for colon cancer histopathological image segmentation.

Recently, Artificial Intelligence namely Deep Learning methods have revolutionized a wide range of domains and applications. Besides, Digital Pathology has so far played a major role in the diagnosis and the prognosis of tumors. However, the characteristics of the Whole Slide Images namely the gigapixel size, high resolution and the shortage of richly labeled samples have hindered the efficiency of classical Machine Learning methods. That goes without saying that traditional methods are poor in generalization to different tasks and data contents. Regarding the success of Deep learning when dealing with Large Scale applications, we have resorted to the use of such models for histopathological image segmentation tasks. First, we review and compare the classical UNet and Att-UNet models for colon cancer WSI segmentation in a sparsely annotated data scenario. Then, we introduce novel enhanced models of the Att-UNet where different schemes are proposed for the skip connections and spatial attention gates positions in the network. In fact, spatial attention gates assist the training process and enable the model to avoid irrelevant feature learning. Alternating the presence of such modules namely in our Alter-AttUNet model adds robustness and ensures better image segmentation results. In order to cope with the lack of richly annotated data in our AiCOLO colon cancer dataset, we suggest the use of a multi-step training strategy that also deals with the WSI sparse annotations and unbalanced class issues. All proposed methods outperform state-of-the-art approaches but Alter-AttUNet generates the best compromise between accurate results and light network. The model achieves 95.88% accuracy with our sparse AiCOLO colon cancer datasets. Finally, to evaluate and validate our proposed architectures we resort to publicly available WSI data: the NCT-CRC-HE-100K, the CRC-5000 and the Warwick colon cancer histopathological dataset. Respective accuracies of 99.65%, 99.73% and 79.03% were reached. A comparison with state-of-art approaches is established to view and compare the key solutions for histopathological image segmentation.

Tumor budding is an independent prognostic factor in stage III colon cancer patients: a post-hoc analysis of the IDEA-France phase III trial (PRODIGE-GERCOR).

BACKGROUND: Histological characteristics at the invasive front may reflect tumor aggressiveness; specifically, tumor budding (Bd) is an emerging prognostic biomarker in colon cancer (CC). We explored further the significance of Bd for risk stratification by evaluating survival of stage III CC patients included in the IDEA-France phase III trial. PATIENTS AND METHODS: This post-hoc study was conducted on tissue slides from 1048 stage III CC patients. Bd was scored by central review by the Bd criteria of the 2016 International Tumor Budding Consensus Conference (ITBCC 2016) and classified as Bd1 (0-4 buds/0.785 mm2), Bd2 (5-9 buds), and Bd3 (≥10 buds) categories. Disease-free survival (DFS) and overall survival (OS) were analyzed by the log-rank test. Clinicopathological features and Immunoscore® were correlated with Bd. RESULTS: Overall, Bd1, Bd2, and Bd3 were observed in 39%, 28%, and 33% of CC, respectively. Bd2 and Bd3 were associated with vascular (P = 0.002) and perineural invasions (P = 0.0009). The 3-year DFS and the 5-year OS rates for Bd (1 versus 2-3) were 79.4% versus 67.2% (P = 0.001) and 89.2% versus 80.8% (P = 0.001), respectively. This was confirmed after adjustment for relevant clinicopathological features for DFS [hazard ratio (HR) 1.41, 95% confidence interval (CI) 1.12-1.77, P = 0.003] and OS (HR 1.65, 95% CI 1.22-2.22, P = 0.001). When combined with pTN stage and Immunoscore® subgroups, Bd significantly improved disease prognostication. CONCLUSIONS: Bd demonstrated its independent prognostic value for DFS and OS. Given these findings, Bd as per the ITBCC 2016 should be mandatory in every pathology report in stage III CC patients. Bd and Immunoscore® could play a complementary role in personalized health care in this setting.

Deep learning for colon cancer histopathological images analysis.

Nowadays, digital pathology plays a major role in the diagnosis and prognosis of tumours. Unfortunately, existing methods remain limited when faced with the high resolution and size of Whole Slide Images (WSIs) coupled with the lack of richly annotated datasets. Regarding the ability of the Deep Learning (DL) methods to cope with the large scale applications, such models seem like an appealing solution for tissue classification and segmentation in histopathological images. This paper focuses on the use of DL architectures to classify and highlight colon cancer regions in a sparsely annotated histopathological data context. First, we review and compare state-of-the-art Convolutional Neural networks (CNN) including the AlexNet, vgg, ResNet, DenseNet and Inception models. To cope with the shortage of rich WSI datasets, we have resorted to the use of transfer learning techniques. This strategy comes with the hallmark of relying on a large size computer vision dataset (ImageNet) to train the network and generate a rich collection of learnt features. The testing and evaluation of such models on our AiCOLO colon cancer dataset ensure accurate patch-level classification results reaching up to 96.98% accuracy rate with ResNet. The CNN models have also been tested and evaluated with the CRC-5000, nct-crc-he-100k and merged datasets. ResNet respectively achieves 96.77%, 99.76% and 99.98% for the three publicly available datasets. Then, we present a pixel-wise segmentation strategy for colon cancer WSIs through the use of both UNet and SegNet models. We introduce a multi-step training strategy as a remedy for the sparse annotation of histopathological images. UNet and SegNet are used and tested in different training scenarios including data augmentation and transfer learning and ensure up to 76.18% and 81.22% accuracy rates. Besides, we test our training strategy and models on the CRC-5000, nct-crc-he-100k and Warwick datasets. Respective accuracy rates of 98.66%, 99.12% and 78.39% were achieved by SegNet. Finally, we analyze the existing models to discover the most suitable network and the most effective training strategy for our colon tumour segmentation case study.1.

Incidental findings in a series of 2500 gene panel tests for a genetic predisposition to cancer: Results and impact on patients.

With next generation sequencing, physicians are faced with more complex and uncertain data, particularly incidental findings (IF). Guidelines for the return of IF have been published by learned societies. However, little is known about how patients are affected by these results in a context of oncogenetic testing. Over 4 years, 2500 patients with an indication for genetic testing underwent a gene cancer panel. If an IF was detected, patients were contacted by a physician/genetic counsellor and invited to take part in a semi-structured interview to assess their understanding of the result, the change in medical care, the psychological impact, and the transmission of results to the family. Fourteen patients (0.56%) were delivered an IF in a cancer predisposition gene (RAD51C, PMS2, SDHC, RET, BRCA2, CHEK2, CDKN2A, CDH1, SUFU). Two patients did not collect the results and another two died before the return of results. Within the 10 patients recontacted, most of them reported surprise at the delivery of IF, but not anxiety. The majority felt they had chosen to obtain the result and enough information to understand it. They all initiated the recommended follow-up and did not regret the procedure. Information regarding the IF was transmitted to their offspring but siblings or second-degree relatives were not consistently informed. No major adverse psychological events were found in our experience. IF will be inherent to the development of sequencing, even for restricted gene panels, so it is important to increase our knowledge on the impact of such results in different contexts.

Inhibition of colon cancer growth by docosahexaenoic acid involves autocrine production of TNFα.

The omega-3 polyunsaturated fatty acid docosahexaenoic acid (DHA) has anti-inflammatory and anti-cancer properties. Among pro-inflammatory mediators, tumor necrosis factor α (TNFα) plays a paradoxical role in cancer biology with induction of cancer cell death or survival depending on the cellular context. The objective of the study was to evaluate the role of TNFα in DHA-mediated tumor growth inhibition and colon cancer cell death. The treatment of human colorectal cancer cells, HCT-116 and HCT-8 cells, with DHA triggered apoptosis in autocrine TNFα-dependent manner. We demonstrated that DHA-induced increased content of TNFα mRNA occurred through a post-transcriptional regulation via the down-regulation of microRNA-21 (miR-21) expression. Treatment with DHA led to nuclear accumulation of Foxo3a that bounds to the miR-21 promoter triggering its transcriptional repression. Moreover, inhibition of RIP1 kinase and AMP-activated protein kinase α reduced Foxo3a nuclear-cytoplasmic shuttling and subsequent increase of TNFα expression through a decrease of miR-21 expression in DHA-treated colon cancer cells. Finally, we were able to show in HCT-116 xenograft tumor-bearing nude mice that a DHA-enriched diet induced a decrease of human miR-21 expression and an increase of human TNFα mRNA expression limiting tumor growth in a cancer cell-derived TNFα dependent manner. Altogether, the present work highlights a novel mechanism for anti-cancer action of DHA involving colon cancer cell death mediated through autocrine action of TNFα.

Liver X receptor ß activation induces pyroptosis of human and murine colon cancer cells.

Liver X receptors (LXRs) have been proposed to have some anticancer properties, through molecular mechanisms that remain elusive. Here we report for the first time that LXR ligands induce caspase-1-dependent cell death of colon cancer cells. Caspase-1 activation requires Nod-like-receptor pyrin domain containing 3 (NLRP3) inflammasome and ATP-mediated P2 × 7 receptor activation. Surprisingly, LXRß is mainly located in the cytoplasm and has a non-genomic role by interacting with pannexin 1 leading to ATP secretion. Finally, LXR ligands have an antitumoral effect in a mouse colon cancer model, dependent on the presence of LXRß, pannexin 1, NLRP3 and caspase-1 within the tumor cells. Our results demonstrate that LXRß, through pannexin 1 interaction, can specifically induce caspase-1-dependent colon cancer cell death by pyroptosis.