Assisted documentation as a new focus for artificial intelligence in endoscopy: the precedent of reliable withdrawal time and image reporting.

BACKGROUND : Reliable documentation is essential for maintaining quality standards in endoscopy; however, in clinical practice, report quality varies. We developed an artificial intelligence (AI)-based prototype for the measurement of withdrawal and intervention times, and automatic photodocumentation. METHOD: A multiclass deep learning algorithm distinguishing different endoscopic image content was trained with 10 557 images (1300 examinations, nine centers, four processors). Consecutively, the algorithm was used to calculate withdrawal time (AI prediction) and extract relevant images. Validation was performed on 100 colonoscopy videos (five centers). The reported and AI-predicted withdrawal times were compared with video-based measurement; photodocumentation was compared for documented polypectomies. RESULTS: Video-based measurement in 100 colonoscopies revealed a median absolute difference of 2.0 minutes between the measured and reported withdrawal times, compared with 0.4 minutes for AI predictions. The original photodocumentation represented the cecum in 88 examinations compared with 98/100 examinations for the AI-generated documentation. For 39/104 polypectomies, the examiners' photographs included the instrument, compared with 68 for the AI images. Lastly, we demonstrated real-time capability (10 colonoscopies). CONCLUSION : Our AI system calculates withdrawal time, provides an image report, and is real-time ready. After further validation, the system may improve standardized reporting, while decreasing the workload created by routine documentation.

Efficacy of zoledronic acid for the elimination of disseminated tumor cells in a clinically relevant, spontaneously metastatic prostate cancer xenograft model.

Bone metastases develop in >90 % of patients with castration-resistant prostate cancer (PCa) through complex interactions between the bone microenvironment and tumor cells. Previous androgen-deprivation therapy (ADT), which is known to cause bone loss, as well as anti-resorptive agents such as zoledronic acid (ZA), used to prevent skeletal complications, may influence these interactions and thereby the growth of disseminated tumor cells (DTC) in the bone marrow (BM). Here, a spontaneously metastatic xenograft tumor model of human PCa was further optimized to mimic the common clinical situation of ADT (castration) combined with primary tumor resection in vivo. The effects of these interventions, alone or in combination with ZA treatment, on tumor cell dissemination to the BM and other distant sites were analyzed. Metastatic burden was quantified by human-specific Alu-qPCR, bioluminescence imaging (BLI), and immunohistochemistry. Further, bone remodeling was assessed by static histomorphometry and serum parameters. Initial comparative analysis between NSG and SCID mice showed that spontaneous systemic dissemination of subcutaneous PC-3 xenograft tumors was considerably enhanced in NSG mice. Primary tumor resection and thereby prolonged observational periods resulted in a higher overall metastatic cell load at necropsy and tumor growth alone caused significant bone loss, which was further augmented by surgical castration. In addition, castrated mice showed a strong trend towards higher bone metastasis loads. Weekly treatment of mice with ZA completely prevented castration- and tumor-induced bone loss but had no effect on bone metastasis burden. Conversely, the total lung metastasis load as determined by BLI was significantly decreased upon ZA treatment. These findings provide a basis for future research on the role of ZA not only in preventing skeletal complications but also in reducing metastasis to other organs.

Quantitative analysis of submucosal excision depth in endoscopic resection for early Barrett's cancer.

BACKGROUND : Following endoscopic resection of early-stage Barrett's esophageal adenocarcinoma (BEA), further oncologic management then fundamentally relies upon the accurate assessment of histopathologic risk criteria, which requires there to be sufficient amounts of submucosal tissue in the resection specimens. METHODS : In 1685 digitized tissue sections from endoscopic mucosal resection (EMR) or endoscopic submucosal dissection (ESD) performed for 76 early BEA cases from three experienced centers, the submucosal thickness was determined, using software developed in-house. Neoplastic lesions were manually annotated. RESULTS : No submucosa was seen in about a third of the entire resection area (mean 33.8 % [SD 17.2 %]), as well as underneath cancers (33.3 % [28.3 %]), with similar results for both resection methods and with respect to submucosal thickness. ESD results showed a greater variability between centers than EMR. In T1b cancers, a higher rate of submucosal defects tended to correlate with R1 resections. CONCLUSION : The absence of submucosa underneath about one third of the tissue of endoscopically resected BEAs should be improved. Results were more center-dependent for ESD than for EMR. Submucosal defects can potentially serve as a parameter for standardized reports.

Possible tumour cell reimplantation during curative endoscopic therapy of superficial Barrett's carcinoma.

BACKGROUND AND AIMS: Endoscopic resection has been established as curative therapy for superficial cancer arising from Barrett's oesophagus (BE); recurrences are very rare. Based on a case series with unusual and massive early recurrences, we analyse the issue of tumour cell reimplantation. METHODS: This hypothesis was developed on the basis of two out of seven patients treated by circumferential (n=6) or nearly circumferential (n=1) en bloc and R0 endoscopic resection of T1 neoplastic BE. Subsequently, a prospective histocytological analysis of endoscope channels and accessories was performed in 2 phases (cytohistological analysis; test for cell viability) in 22 different oesophageal carcinoma patients undergoing endoscopy. Finally, cultures from two oesophageal adenocarcinoma cell lines were incubated with different triamcinolone concentrations (0.625-10 mg/mL); cell growth was determined on a Multiwell plate reader. RESULTS: Cancer regrowth in the two suspicious cases (male, 78/71 years) occurred 7 and 1 months, respectively, after curative tumour resection. Subsequent surgery showed advanced tumours (T2) with lymph node metastases; one patient died. On cytohistological examinations of channels and accessories, suspicious/neoplastic cells were found in 4/10 superficial and in all 5 advanced cancers. Further analyses in seven further advanced adenocarcinoma cases showed viable cells in two channel washing specimens. Finally, cell culture experiments demonstrated enhanced tumour cell growth by triamcinolone after 24 hours compared with controls. CONCLUSIONS: Tumour cell reimplanation from contaminated endoscopes and accessories is a possible cause of local recurrence after curative endoscopic therapy for superficial Barrett carcinoma; also, corticosteroid injection could have promoted tumour regrowth in these cases.

Multi-scale fully convolutional neural networks for histopathology image segmentation: From nuclear aberrations to the global tissue architecture.

Histopathologic diagnosis relies on simultaneous integration of information from a broad range of scales, ranging from nuclear aberrations (≈O(0.1µm)) through cellular structures (≈O(10µm)) to the global tissue architecture (⪆O(1mm)). To explicitly mimic how human pathologists combine multi-scale information, we introduce a family of multi-encoder fully-convolutional neural networks with deep fusion. We present a simple block for merging model paths with differing spatial scales in a spatial relationship-preserving fashion, which can readily be included in standard encoder-decoder networks. Additionally, a context classification gate block is proposed as an alternative for the incorporation of global context. Our experiments were performed on three publicly available whole-slide images of recent challenges (PAIP 2019: hepatocellular carcinoma segmentation; BACH 2020: breast cancer segmentation; CAMELYON 2016: metastasis detection in lymph nodes). The multi-scale architectures consistently outperformed the baseline single-scale U-Nets by a large margin. They benefit from local as well as global context and particularly a combination of both. If feature maps from different scales are fused, doing so in a manner preserving spatial relationships was found to be beneficial. Deep guidance by a context classification loss appeared to improve model training at low computational costs. All multi-scale models had a reduced GPU memory footprint compared to ensembles of individual U-Nets trained on different image scales. Additional path fusions were shown to be possible at low computational cost, opening up possibilities for further, systematic and task-specific architecture optimisation. The findings demonstrate the potential of the presented family of human-inspired, end-to-end trainable, multi-scale multi-encoder fully-convolutional neural networks to improve deep histopathologic diagnosis by extensive integration of largely different spatial scales.

PAIP 2019: Liver cancer segmentation challenge.

Pathology Artificial Intelligence Platform (PAIP) is a free research platform in support of pathological artificial intelligence (AI). The main goal of the platform is to construct a high-quality pathology learning data set that will allow greater accessibility. The PAIP Liver Cancer Segmentation Challenge, organized in conjunction with the Medical Image Computing and Computer Assisted Intervention Society (MICCAI 2019), is the first image analysis challenge to apply PAIP datasets. The goal of the challenge was to evaluate new and existing algorithms for automated detection of liver cancer in whole-slide images (WSIs). Additionally, the PAIP of this year attempted to address potential future problems of AI applicability in clinical settings. In the challenge, participants were asked to use analytical data and statistical metrics to evaluate the performance of automated algorithms in two different tasks. The participants were given the two different tasks: Task 1 involved investigating Liver Cancer Segmentation and Task 2 involved investigating Viable Tumor Burden Estimation. There was a strong correlation between high performance of teams on both tasks, in which teams that performed well on Task 1 also performed well on Task 2. After evaluation, we summarized the top 11 team's algorithms. We then gave pathological implications on the easily predicted images for cancer segmentation and the challenging images for viable tumor burden estimation. Out of the 231 participants of the PAIP challenge datasets, a total of 64 were submitted from 28 team participants. The submitted algorithms predicted the automatic segmentation on the liver cancer with WSIs to an accuracy of a score estimation of 0.78. The PAIP challenge was created in an effort to combat the lack of research that has been done to address Liver cancer using digital pathology. It remains unclear of how the applicability of AI algorithms created during the challenge can affect clinical diagnoses. However, the results of this dataset and evaluation metric provided has the potential to aid the development and benchmarking of cancer diagnosis and segmentation.

Digital Health meets Hamburg integrated medical degree program iMED: concept and introduction of the new interdisciplinary 2nd track Digital Health.

Digitalization in medicine is transforming the everyday work and the environment of current and future physicians - and thereby brings new competencies required by the medical profession. The necessity for a curricular integration of related digital medicine and, in more general, digital health topics is mostly undisputed; however, few specific concepts and experience reports are available. Therefore, the present article reports on the aims, the implementation, and the initial experiences of the integration of the topic Digital Health as a longitudinal elective course (2nd track) into the integrated medical degree program iMED in Hamburg.

Skin Lesion Classification Using CNNs With Patch-Based Attention and Diagnosis-Guided Loss Weighting.

OBJECTIVE: This paper addresses two key problems of skin lesion classification. The first problem is the effective use of high-resolution images with pretrained standard architectures for image classification. The second problem is the high-class imbalance encountered in real-world multi-class datasets. METHODS: To use high-resolution images, we propose a novel patch-based attention architecture that provides global context between small, high-resolution patches. We modify three pretrained architectures and study the performance of patch-based attention. To counter class imbalance problems, we compare oversampling, balanced batch sampling, and class-specific loss weighting. Additionally, we propose a novel diagnosis-guided loss weighting method that takes the method used for ground-truth annotation into account. RESULTS: Our patch-based attention mechanism outperforms previous methods and improves the mean sensitivity by [Formula: see text]. Class balancing significantly improves the mean sensitivity and we show that our diagnosis-guided loss weighting method improves the mean sensitivity by [Formula: see text] over normal loss balancing. CONCLUSION: The novel patch-based attention mechanism can be integrated into pretrained architectures and provides global context between local patches while outperforming other patch-based methods. Hence, pretrained architectures can be readily used with high-resolution images without downsampling. The new diagnosis-guided loss weighting method outperforms other methods and allows for effective training when facing class imbalance. SIGNIFICANCE: The proposed methods improve automatic skin lesion classification. They can be extended to other clinical applications where high-resolution image data and class imbalance are relevant.

[Artificial Intelligence in Endoscopy: Deep Neural Nets for Endoscopic Computer Vision - Methods & Perspectives]. / Künstliche Intelligenz in der Endoskopie: Neuronale Netze und maschinelles Sehen ­ Techniken und Perspektiven.

Artificial neural networks, as a specific approach towards artificial intelligence (AI), can open up a variety of new perspectives for endoscopy, such as automated lesion detection and the precise prediction of a lesion's histology by its endoscopic appearance. Whilst early experiments do suggest an enormous potential for these methods, public expectations on their application in various fields of medicine sometimes appear to be grounded on general fascination rather than detailed understanding of their inner workings. Based on a selective review of the literature, this article shall convey an intuitive understanding of the underlying methods in order to help close the gap between functioning and fascination and allow for a realistic discussion of their perspectives and limitations in endoscopy.After decades of research, the success of deep neuronal networks in image classification has provoked rising interest for AI during recent years. We quickly touch upon the developments surrounding this breakthrough and the reasons for their impact on various disciplines much beyond computer science. Through a comparison with the functioning of the human vision system, we aim to understand the mechanisms of these techniques and their success in computer vision tasks in detail. Based on these considerations, we analyse the functioning of some important AI applications in endoscopy, deduce specific limitations and perspectives, discuss the current state of their evaluation in practical endoscopy and make a plea for the need for additional and realistic tests. Moreover, we seek to give an impression of some further specific applications that can currently be foreseen and how these can shape the role that AI might finally acquire in the routine clinical practice of GI endoscopy.

Modeling Growth of Tumors and Their Spreading Behavior Using Mathematical Functions.

Computer simulations of the spread of malignant tumor cells in an entire organism provide important insights into the mechanisms of metastatic progression. Key elements for the usefulness of these models are the adequate selection of appropriate mathematical models describing the tumor growth and its parametrization as well as a proper choice of the fractal dimension of the blood vessels in the primary tumor. In addition, survival in the bloodstream and evasion into the connective spaces of the target organ of the future metastasis have to be modeled. Determination of these from experimental models is complicated by systematic and unsystematic experimental errors which are difficult to assess. In this chapter, we demonstrate how to select the best-suited mathematical function to describe tumor growth for experimental xenograft mouse tumor models and how to parametrize them. Common pitfalls and problems are described as well as methods to avoid them.

Locally Ablative Radiation Therapy of a Primary Human Small Cell Lung Cancer Tumor Decreases the Number of Spontaneous Metastases in Two Xenograft Models.

PURPOSE: To investigated the influence of radiation therapy (RT), surgery (OP), radio-chemotherapy (RChT), or chemotherapy (ChT) on small cell lung cancer metastases in 2 xenograft models. METHODS AND MATERIALS: A total of 1 × 106 human small cell lung cancer cells (OH1, H69) were subcutaneously injected into severe combined immunodeficiency mice to form a local primary tumor node at the lower trunk. Radiation therapy, OP, RChT, or ChT were started after development of palpable tumors. Chemotherapy was given as a single intraperitoneal injection of cisplatin. Radiation therapy was 5 × 10 Gy on the local tumor node. Two additional groups were implemented to assess primary tumors and distant metastases in untreated mice at the beginning (control group A) and at the end of the experiment (control group B). Proapoptotic, antiproliferative, antiangiogenic, and hypoxic effects were assessed by Feulgen, Ki67, S1P1 receptor, and hypoxia-inducible factor 1α staining, respectively. Quantitative Alu-polymerase chain reaction was used to determine circulating tumor cells in the blood, and disseminated tumor cells in the lungs, bone marrow, liver, and brain. RESULTS: In both xenograft models, RT and RChT abrogated local tumor growth, indicated by increased apoptosis, decreased cell proliferation, and reduced microvessel density (equally affecting vessels of all diameters). Regarding metastases, RT and RChT not only counteracted the time-dependent increase of dissemination but also decreased the metastatic load pre-existing at therapy induction in the blood, lungs, and liver. Only in the case of relapse-free surgery could similar effects be achieved by OP. CONCLUSIONS: Our models provide evidence that RT and RChT ablate the primary tumor and inhibit metastasis development over time. Upon local recurrence, RT showed beneficial effects compared with OP with regard to suppression of circulating tumor cells and disseminated tumor cells.

Radiotherapy and chemotherapy change vessel tree geometry and metastatic spread in a small cell lung cancer xenograft mouse tumor model.

BACKGROUND: Tumor vasculature is critical for tumor growth, formation of distant metastases and efficiency of radio- and chemotherapy treatments. However, how the vasculature itself is affected during cancer treatment regarding to the metastatic behavior has not been thoroughly investigated. Therefore, the aim of this study was to analyze the influence of hypofractionated radiotherapy and cisplatin chemotherapy on vessel tree geometry and metastasis formation in a small cell lung cancer xenograft mouse tumor model to investigate the spread of malignant cells during different treatments modalities. METHODS: The biological data gained during these experiments were fed into our previously developed computer model "Cancer and Treatment Simulation Tool" (CaTSiT) to model the growth of the primary tumor, its metastatic deposit and also the influence on different therapies. Furthermore, we performed quantitative histology analyses to verify our predictions in xenograft mouse tumor model. RESULTS: According to the computer simulation the number of cells engrafting must vary considerably to explain the different weights of the primary tumor at the end of the experiment. Once a primary tumor is established, the fractal dimension of its vasculature correlates with the tumor size. Furthermore, the fractal dimension of the tumor vasculature changes during treatment, indicating that the therapy affects the blood vessels' geometry. We corroborated these findings with a quantitative histological analysis showing that the blood vessel density is depleted during radiotherapy and cisplatin chemotherapy. The CaTSiT computer model reveals that chemotherapy influences the tumor's therapeutic susceptibility and its metastatic spreading behavior. CONCLUSION: Using a system biological approach in combination with xenograft models and computer simulations revealed that the usage of chemotherapy and radiation therapy determines the spreading behavior by changing the blood vessel geometry of the primary tumor.