Influence of artificial intelligence on the diagnostic performance of endoscopists in the assessment of Barrett's esophagus: a tandem randomized and video trial

BACKGROUND

This study evaluated the effect of an artificial intelligence (AI)-based clinical decision support system on the performance and diagnostic confidence of endoscopists in their assessment of Barrett's esophagus (BE).

METHODS

96 standardized endoscopy videos were assessed by 22 endoscopists with varying degrees of BE experience from 12 centers. Assessment was randomized into two video sets: group A (review first without AI and second with AI) and group B (review first with AI and second without AI). Endoscopists were required to evaluate each video for the presence of Barrett's esophagus-related neoplasia (BERN) and then decide on a spot for a targeted biopsy. After the second assessment, they were allowed to change their clinical decision and confidence level.

RESULTS

AI had a stand-alone sensitivity, specificity, and accuracy of 92.2%, 68.9%, and 81.3%, respectively. Without AI, BE experts had an overall sensitivity, specificity, and accuracy of 83.3%, 58.1%, and 71.5%, respectively. With AI, BE nonexperts showed a significant improvement in sensitivity and specificity when videos were assessed a second time with AI (sensitivity 69.8% [95%CI 65.2%-74.2%] to 78.0% [95%CI 74.0%-82.0%]; specificity 67.3% [95%CI 62.5%-72.2%] to 72.7% [95%CI 68.2%-77.3%]). In addition, the diagnostic confidence of BE nonexperts improved significantly with AI.

CONCLUSION

BE nonexperts benefitted significantly from additional AI. BE experts and nonexperts remained significantly below the stand-alone performance of AI, suggesting that there may be other factors influencing endoscopists' decisions to follow or discard AI advice.

Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation

Semantic segmentation is an essential task in medical imaging research. Many powerful deep-learning-based approaches can be employed for this problem, but they are dependent on the availability of an expansive labeled dataset. In this work, we augment such supervised segmentation models to be suitable for learning from unlabeled data. Our semi-supervised approach, termed Error-Correcting Mean-Teacher, uses an exponential moving average model like the original Mean Teacher but introduces our new paradigm of error correction. The original segmentation network is augmented to handle this secondary correction task. Both tasks build upon the core feature extraction layers of the model. For the correction task, features detected in the input image are fused with features detected in the predicted segmentation and further processed with task-specific decoder layers. The combination of image and segmentation features allows the model to correct present mistakes in the given input pair. The correction task is trained jointly on the labeled data. On unlabeled data, the exponential moving average of the original network corrects the student's prediction. The combined outputs of the students' prediction with the teachers' correction form the basis for the semi-supervised update. We evaluate our method with the 2017 and 2018 Robotic Scene Segmentation data, the ISIC 2017 and the BraTS 2020 Challenges, a proprietary Endoscopic Submucosal Dissection dataset, Cityscapes, and Pascal VOC 2012. Additionally, we analyze the impact of the individual components and examine the behavior when the amount of labeled data varies, with experiments performed on two distinct segmentation architectures. Our method shows improvements in terms of the mean Intersection over Union over the supervised baseline and competing methods. Code is available at https://github.com/CloneRob/ECMT.

Vessel and tissue recognition during third-space endoscopy using a deep learning algorithm

In this study, we aimed to develop an artificial intelligence clinical decision support solution to mitigate operator-dependent limitations during complex endoscopic procedures such as endoscopic submucosal dissection and peroral endoscopic myotomy, for example, bleeding and perforation. A DeepLabv3-based model was trained to delineate vessels, tissue structures and instruments on endoscopic still images from such procedures. The mean cross-validated Intersection over Union and Dice Score were 63% and 76%, respectively. Applied to standardised video clips from third-space endoscopic procedures, the algorithm showed a mean vessel detection rate of 85% with a false-positive rate of 0.75/min. These performance statistics suggest a potential clinical benefit for procedure safety, time and also training.

Endoscopic prediction of submucosal invasion in Barrett's cancer with the use of artificial intelligence: a pilot study

BACKGROUND

The accurate differentiation between T1a and T1b Barrett's-related cancer has both therapeutic and prognostic implications but is challenging even for experienced physicians. We trained an artificial intelligence (AI) system on the basis of deep artificial neural networks (deep learning) to differentiate between T1a and T1b Barrett's cancer on white-light images.

METHODS

Endoscopic images from three tertiary care centers in Germany were collected retrospectively. A deep learning system was trained and tested using the principles of cross validation. A total of 230 white-light endoscopic images (108 T1a and 122 T1b) were evaluated using the AI system. For comparison, the images were also classified by experts specialized in endoscopic diagnosis and treatment of Barrett's cancer.

RESULTS

The sensitivity, specificity, F1 score, and accuracy of the AI system in the differentiation between T1a and T1b cancer lesions was 0.77, 0.64, 0.74, and 0.71, respectively. There was no statistically significant difference between the performance of the AI system and that of experts, who showed sensitivity, specificity, F1, and accuracy of 0.63, 0.78, 0.67, and 0.70, respectively.

CONCLUSION

This pilot study demonstrates the first multicenter application of an AI-based system in the prediction of submucosal invasion in endoscopic images of Barrett's cancer. AI scored equally to international experts in the field, but more work is necessary to improve the system and apply it to video sequences and real-life settings. Nevertheless, the correct prediction of submucosal invasion in Barrett's cancer remains challenging for both experts and AI.

Convolutional Neural Networks for the evaluation of cancer in Barrett's esophagus: Explainable AI to lighten up the black-box

Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their level of accountability and transparency must be provided in such evaluations. The reliability related to machine learning predictions must be explained and interpreted, especially if diagnosis support is addressed. For this task, the black-box nature of deep learning techniques must be lightened up to transfer its promising results into clinical practice. Hence, we aim to investigate the use of explainable artificial intelligence techniques to quantitatively highlight discriminative regions during the classification of early-cancerous tissues in Barrett's esophagus-diagnosed patients. Four Convolutional Neural Network models (AlexNet, SqueezeNet, ResNet50, and VGG16) were analyzed using five different interpretation techniques (saliency, guided backpropagation, integrated gradients, input × gradients, and DeepLIFT) to compare their agreement with experts' previous annotations of cancerous tissue. We could show that saliency attributes match best with the manual experts' delineations. Moreover, there is moderate to high correlation between the sensitivity of a model and the human-and-computer agreement. The results also lightened that the higher the model's sensitivity, the stronger the correlation of human and computational segmentation agreement. We observed a relevant relation between computational learning and experts' insights, demonstrating how human knowledge may influence the correct computational learning.

Assisting Barrett's esophagus identification using endoscopic data augmentation based on Generative Adversarial Networks

Barrett's esophagus figured a swift rise in the number of cases in the past years. Although traditional diagnosis methods offered a vital role in early-stage treatment, they are generally time- and resource-consuming. In this context, computer-aided approaches for automatic diagnosis emerged in the literature since early detection is intrinsically related to remission probabilities. However, they still suffer from drawbacks because of the lack of available data for machine learning purposes, thus implying reduced recognition rates. This work introduces Generative Adversarial Networks to generate high-quality endoscopic images, thereby identifying Barrett's esophagus and adenocarcinoma more precisely. Further, Convolution Neural Networks are used for feature extraction and classification purposes. The proposed approach is validated over two datasets of endoscopic images, with the experiments conducted over the full and patch-split images. The application of Deep Convolutional Generative Adversarial Networks for the data augmentation step and LeNet-5 and AlexNet for the classification step allowed us to validate the proposed methodology over an extensive set of datasets (based on original and augmented sets), reaching results of 90% of accuracy for the patch-based approach and 85% for the image-based approach. Both results are based on augmented datasets and are statistically different from the ones obtained in the original datasets of the same kind. Moreover, the impact of data augmentation was evaluated in the context of image description and classification, and the results obtained using synthetic images outperformed the ones over the original datasets, as well as other recent approaches from the literature. Such results suggest promising insights related to the importance of proper data for the accurate classification concerning computer-assisted Barrett's esophagus and adenocarcinoma detection.

A technical review of artificial intelligence as applied to gastrointestinal endoscopy: clarifying the terminology

Background and aim  The growing number of publications on the application of artificial intelligence (AI) in medicine underlines the enormous importance and potential of this emerging field of research. In gastrointestinal endoscopy, AI has been applied to all segments of the gastrointestinal tract most importantly in the detection and characterization of colorectal polyps. However, AI research has been published also in the stomach and esophagus for both neoplastic and non-neoplastic disorders. The various technical as well as medical aspects of AI, however, remain confusing especially for non-expert physicians. This physician-engineer co-authored review explains the basic technical aspects of AI and provides a comprehensive overview of recent publications on AI in gastrointestinal endoscopy. Finally, a basic insight is offered into understanding publications on AI in gastrointestinal endoscopy.

A survey on Barrett's esophagus analysis using machine learning

This work presents a systematic review concerning recent studies and technologies of machine learning for Barrett's esophagus (BE) diagnosis and treatment. The use of artificial intelligence is a brand new and promising way to evaluate such disease. We compile some works published at some well-established databases, such as Science Direct, IEEEXplore, PubMed, Plos One, Multidisciplinary Digital Publishing Institute (MDPI), Association for Computing Machinery (ACM), Springer, and Hindawi Publishing Corporation. Each selected work has been analyzed to present its objective, methodology, and results. The BE progression to dysplasia or adenocarcinoma shows a complex pattern to be detected during endoscopic surveillance. Therefore, it is valuable to assist its diagnosis and automatic identification using computer analysis. The evaluation of the BE dysplasia can be performed through manual or automated segmentation through machine learning techniques. Finally, in this survey, we reviewed recent studies focused on the automatic detection of the neoplastic region for classification purposes using machine learning methods.

What can patients do to facilitate shared decision making? A qualitative study of patients with depression or schizophrenia and psychiatrists

PURPOSE

Patient involvement in decision making is endorsed by patients and professionals. While research has recently been conducted on how professionals can promote shared decision making (SDM), little is known about how patients can also facilitate SDM.

METHODS

Seven focus groups were conducted: 3 with psychiatrists and 4 with patients with schizophrenia or depression. The focus groups were transcribed and independently coded line by line by 2 researchers. Data were analyzed using content analysis.

RESULTS

Seven themes related to patient attitudes and behaviors were identified: honesty and openness with one's psychiatrist and oneself, trust in one's psychiatrist and patience with the treatment, respect and politeness, informing the psychiatrist and giving feedback, engagement/active participation during the consultation, gathering information/preparing for the consultation and implementing decisions. Barriers (e.g., avolition, lack of decisional capacity, powerlessness during involuntary treatment) and facilitators of active patient behavior were also identified.

CONCLUSIONS

There are various ways in which patients can facilitate SDM/play a more active role in decision making, with patients emphasizing being open and honest and psychiatrists emphasizing being active in the consultation. Interventions to increase active patient behavior may enhance SDM in mental health care.

[Understanding the term burnout in psychiatry and psychotherapy]

BACKGROUND

Currently there is much debate about the concept of burnout and its use as a diagnostic entity. The aim of the present survey was to present the view of mental health professionals towards the concept of burnout.

METHODS

A total of 300 mental health professionals were surveyed using a structured questionnaire.

RESULTS

The majority of participants see burnout as a state of exhaustion which constitutes a risk factor for later developing a mental disorder. Participants reported that from their point of view typical triggers for burnout exist while symptoms overlap to a great extent with depression. Psychotherapy as well as interventions at the workplace are regarded as promising interventions; however, in the clinical routine only a minority of participants actually contacted the patients' workplace. In the participants workplace settings most Burnout-Patients suffered from a diagnosis defined in ICD 10 but judged themselves to be suffering from burnout.

DISCUSSION

Burnout-Patients in mental health settings differ from the picture currently drawn in the media, probably because Burnout-Patients reach the mental health sector only after already having developed a manifest psychiatric disorder.

Confirmation bias: why psychiatrists stick to wrong preliminary diagnoses

BACKGROUND

Diagnostic errors can have tremendous consequences because they can result in a fatal chain of wrong decisions. Experts assume that physicians' desire to confirm a preliminary diagnosis while failing to seek contradictory evidence is an important reason for wrong diagnoses. This tendency is called 'confirmation bias'.

METHOD

To study whether psychiatrists and medical students are prone to confirmation bias and whether confirmation bias leads to poor diagnostic accuracy in psychiatry, we presented an experimental decision task to 75 psychiatrists and 75 medical students.

RESULTS

A total of 13% of psychiatrists and 25% of students showed confirmation bias when searching for new information after having made a preliminary diagnosis. Participants conducting a confirmatory information search were significantly less likely to make the correct diagnosis compared to participants searching in a disconfirmatory or balanced way [multiple logistic regression: odds ratio (OR) 7.3, 95% confidence interval (CI) 2.53-21.22, p<0.001; OR 3.2, 95% CI 1.23-8.56, p=0.02]. Psychiatrists conducting a confirmatory search made a wrong diagnosis in 70% of the cases compared to 27% or 47% for a disconfirmatory or balanced information search (students: 63, 26 and 27%). Participants choosing the wrong diagnosis also prescribed different treatment options compared with participants choosing the correct diagnosis.

CONCLUSIONS

Confirmatory information search harbors the risk of wrong diagnostic decisions. Psychiatrists should be aware of confirmation bias and instructed in techniques to reduce bias.

Psychiatrists' attitude to antipsychotic depot treatment in patients with first-episode schizophrenia

OBJECTIVE

Despite good clinical evidence, depot antipsychotics are only seldom prescribed to patients with first episode schizophrenia. The present study aims at investigating psychiatrists' reasons for this reservation.

METHOD

We surveyed 198 psychiatrists on their attitude toward offering depot treatment to first episode patients (FEP). Participants scored the extent of influence of individual factors on their decision on a seven-point-scale, additional data on their prescription practice and estimation of the relapse risk of FEP were collected.

RESULTS

Psychiatrists reported that only three out of 12 factors were of influence. These were the limited availability of different second generation antipsychotic depot drugs, the frequent rejection of the depot offer by the patients and the patients' skepticism based on the lack in experience of a relapse.

CONCLUSIONS

There is actually little specific reason for not prescribing depot to FEP according to the current survey. For those factors being reported to be of influence, psychoeducation, including profound information on depot treatment, the development of additional SGA depot drugs and the standard offer of depot treatment to all FEP in a shared-decision-making may be considered.

How psychiatrists inform themselves and their patients about risks and benefits of antipsychotic treatment

OBJECTIVE

In order to choose the best treatment option, physicians have to inform themselves and their patients about both the benefits and risks of available treatment options equally. Our study aims to investigate whether psychiatrists actually do conduct such a balanced information search and presentation.

METHOD

Psychiatrists' information search and information presentation to a patient with schizophrenia were studied using two separate experiments. In both, participants were presented with hypothetical case vignettes and descriptions of fictitious antipsychotics.

RESULTS

When searching for information, psychiatrists looked more for risks than benefits of antipsychotic treatment options (t = -3.4, df = 74, P = 0.001). However, when informing a patient, they named more benefits than risks (t = 17.1, df = 224, P < 0.001).

CONCLUSION

The risk-biased information search presumably follows the principle of 'primum non nocere'. The benefit-biased information presentation might be motivated by the wish to persuade patients to accept the proposed therapy.

Discordant prevalence of chlamydia trachomatis in asymptomatic couples screened using urine ligase chain reaction

The following study was conducted to determine whether there would be an effect on the prevalence of Chlamydia trachomatis if both partners in a sexual relationship, rather than only one, underwent screening. First-void urine samples were collected from 1,690 asymptomatic women (mean age, 30 years; range, 15-70 years) and their male sex partners (mean age, 33 years; range, 16-71 years). The duration of sexual partnership for these subjects ranged from 2 months to more than 10 years.. At the time of testing, 687 of the women were pregnant. Ligase chain reaction testing revealed that 42 (2.5%) female and 63 (3.7%) male urine samples were positive. Detection rates for Chlamydia trachomatis differed for males and females, a difference that was found to be significant (P<0.0046, McNemar chi-square). Both partners tested positive in 27 (1.6%) couples, whereas at least one partner tested positive in 78 (4.6%) couples. Thus, screening males for Chlamydia trachomatis would have identified 63 (81%) of these 78 couples compared with only 42 (54%) couples had females been screened exclusively. In standard clinical practice, women most often undergo screening. The results of this study underscore the need to screen both males and females for Chlamydia trachomatis.

The Chlamydomonas reinhardtii MoCo carrier protein is multimeric and stabilizes molybdopterin cofactor in a molybdate charged form

In Chlamydomonas reinhardtii, molybdopterin cofactor (MoCo) able to reconstitute active nitrate reductase (NR) with apoenzyme from the Neurospora crassa mutant nit-1 was found mostly bound to a carrier protein (CP). This protein is scarce in the algal free extracts and has been purified 520-fold. MoCoCP is a protein of 64 kDa with subunits of 16.5 kDa and an isoelectric point of 4.5. In contrast to free MoCo, MoCo bound to CP was remarkably protected against inactivation under both aerobic conditions and basic pH. MocoCP transferred active MoCo to apoNR in vitro without addition of molybdate, though reconstituted activity was 20% higher in the presence of molybdate. Incubation with tungstate specifically inhibited MoCoCP activity but had no effect on the activity of free MoCo released from milk xanthine oxidase. MoCoCP did not charge molybdate unless in the presence of N. crassa extracts. Our data support that MoCoCP stabilizes MoCo in an active form charged with molybdate to provide MoCo to apomolybdoenzymes.

Intron-dependent transient expression of the maize GapA1 gene

Transient expression experiments show that the maize GapA1 promoter exhibits a requirement for sequences contained within intron 1 and surrounding exon border regions for expression in maize Black Mexican Sweet cells. Maize GapA1-promoter constructs lacking intron 1 are inactive. Intron 1 and its exon border sequences, when reintroduced into constructs lacking introns, restore gene activity whereas intron 2 and its exon borders to not. The minimal promoter so defined encompasses roughly 250 bp upstream of the in vivo transcription start and appears also to include intron 1. An octameric sequence was identified in intron 1 of maize GapA1 which is similar to sequence motifs found in other maize introns known to increase transient expression. Partial restoration of gene expression in GapA1 constructs lacking intron 1 was achieved through insertion of the identified octameric sequence.

[Highly differentiated adenocarcinoma in a villous adenoma of the common bile duct near the papilla]

A 65-year-old man, in hospital for treatment for bladder carcinoma, was by chance found also to have a so far asymptomatic obstructive jaundice with hydrops of the gallbladder. Endoscopic retrograde cholangiography revealed, in addition to a papilla distorted by tumour, a clearly obstructed choledochal duct with dilated intrahepatic bile ducts, caused by a 2-3 cm prepapillary intraductal choledochus tumour, which in the biopsy specimen corresponded to papillomatous structures with occasional medium-grade epithelial dysplasias. The patients, who initially declined operation, was five weeks later re-admitted with the clinical picture of an acute abdomen. In view of the history the surgeon performed a partial duodenopancreatectomy despite the acute emergency. Histology revealed a highly differentiated adenocarcinoma within a villous adenoma of the choledochal duct near the papilla.

A binary vector for transferring genomic libraries to plants

The transformation of mutant plants with a complete recombinant library derived from wild-type DNA followed by assay of transformed plants for complementation of the mutant phenotype is a promising method for the isolation of plant genes. The small genome of Arabidopsis thaliana is a good candidate for attempting this so-called shotgun transformation. We present the properties of an A. thaliana genomic library cloned in a binary vector, pC22. This vector, designed to introduce genomic libraries into plants, contains the oriV of the Ri plasmid pRiHR1 by which it replicates perfectly stably in Agrobacterium. Upon transfer of the library from E. coli to A. tumefaciens large differences in transfer efficiencies of individual recombinant clones were observed. There is a direct relation between transfer efficiency and stability of the recombinant clones both in E. coli and A. tumefaciens. The stability is independent of the insert size, but seems to be related to the nature of the insert DNA. The feasibility of shotgun transformation and problems of statistical sampling are discussed.

[2 cases of primary sclerosing cholangitis--the ERCP documentation of a 3-year course of immunosuppressive therapy]

X-ray characteristics of two cases of primary sclerosing cholangitis are described. A discussion of clinical, immunological and histological features of the disease and the 3-years progress of a 40 year old man during treatment with immunesuppressive therapy (Imurek at the beginning with cortison) is shown. The stop of the progression of the disease is supposed.