Refine
Document Type
- Conference proceeding (33)
- Journal article (29)
- Book chapter (1)
Is part of the Bibliography
- yes (63)
Institute
- Informatik (63)
Publisher
- Springer (16)
- De Gruyter (11)
- Deutsche Gesellschaft für Computer- und Roboterassistierte Chirurgie e.V. (7)
- SPIE. The International Society for Optical Engineering (5)
- IEEE (4)
- IOP Publishing (4)
- Deutsche Gesellschaft für Computer- und Roboterassistierte Chirurgie e. V. (3)
- GMDS e.V. (2)
- Thieme (2)
- Elsevier (1)
- Morressier (1)
- NextMed (1)
- Taylor & Francis (1)
- libreriauniversitaria.it.edizioni (1)
Die DGCH registriert vermehrt Klagen aus der klinischen Praxis hinsichtlich der nicht vollständigen Vernetzung bzw. Integration von Gerätesystemen im Chirurgischen OP. Die Anzahl, der Funktionsumfang und der Komplexitätsgrad der verwendeten Geräte nehmen ständig zu und machen die Bedienung immer aufwendiger und damit schwieriger und fehleranfälliger, sodass eine Verbesserung bei der Unterstützung im Ablauf wünschenswert ist. Die Sektion Computer- und telematikassistierte Chirurgie (CTAC) der DGCH hat es auf Veranlassung des Generalsekretärs deshalb übernommen, eine aktuelle Bestandsaufnahme vorzunehmen und mögliche Ansätze zur Verbesserung des derzeitigen Status zu bewerten.
Purpose
Artificial intelligence (AI), in particular deep learning (DL), has achieved remarkable results for medical image analysis in several applications. Yet the lack of human-like explanations of such systems is considered the principal restriction before utilizing these methods in clinical practice (Yang, Ye, & Xia, 2022).
Methods
Explainable Artificial Intelligence (XAI) provides a human-explainable and interpretable description of the “black-box” nature of DL (Gulum, Trombley, & Kantardzic, 2021). An effective XAI diagnosis generator, namely NeuroXAI (refer to Fig. 1), has been developed to extract 3D explanations from convolutional neural networks (CNN) models of brain gliomas (Zeineldin et al., 2022). By providing visual justification maps, NeuroXAI can help make DL models transparent and thus increase the trust of medical experts.
Results
NeuroXAI has been applied to two applications of the most widely investigated problems in brain imaging analysis, i.e. image classification and segmentation using magnetic resonance imaging (MRI). Visual attention maps of multiple XAI methods have been generated and compared for both applications, which could help to provide transparency about the performance of DL systems.
Conclusion
NeuroXAI helps to understand the prediction process of 3D CNN networks for brain glioma using human-understandable explanations. Results revealed that the investigated DL models behave in a logical human-like manner and can improve the analytical process of the MRI images systematically. Due to its open architecture, ease of implementation, and scalability to new XAI methods, NeuroXAI could be utilized to assist medical professionals in the detection and diagnosis of brain tumors. NeuroXAI code is publicly accessible at https://github.com/razeineldin/NeuroXAI
Purpose
Artificial intelligence (AI), in particular deep neural networks, has achieved remarkable results for medical image analysis in several applications. Yet the lack of explainability of deep neural models is considered the principal restriction before applying these methods in clinical practice.
Methods
In this study, we propose a NeuroXAI framework for explainable AI of deep learning networks to increase the trust of medical experts. NeuroXAI implements seven state-of-the-art explanation methods providing visualization maps to help make deep learning models transparent.
Results
NeuroXAI has been applied to two applications of the most widely investigated problems in brain imaging analysis, i.e., image classification and segmentation using magnetic resonance (MR) modality. Visual attention maps of multiple XAI methods have been generated and compared for both applications. Another experiment demonstrated that NeuroXAI can provide information flow visualization on internal layers of a segmentation CNN.
Conclusion
Due to its open architecture, ease of implementation, and scalability to new XAI methods, NeuroXAI could be utilized to assist radiologists and medical professionals in the detection and diagnosis of brain tumors in the clinical routine of cancer patients. The code of NeuroXAI is publicly accessible at https://github.com/razeineldin/NeuroXAI.
Hintergrund: Endoskopische Operationsverfahren haben sich als Goldstandard in der Nasennebenhöhlen-(NNH-)Chirurgie etabliert. Den sich daraus ergebenden Herausforderungen für die chirurgische Ausbildung kann durch den Einsatz von Virtuelle-Realität-(VR-)Trainingssimulatoren begegnet werden. Bislang wurde eine Reihe von Simulatoren für NNH-Operationen entwickelt. Frühere Studien im Hinblick auf den Trainingseffekt wurden jedoch nur mit medizinisch vorgebildeten Probanden durchgeführt oder es wurde nicht über dessen zeitlichen Verlauf berichtet.
Methoden: Ein NNH-CT-Datensatz wurde nach der Segmentierung in ein 3-dimensionales, polygonales Oberflächenmodell überführt und mithilfe von originalem Fotomaterial texturiert. Die Interaktion mit der virtuellen Umgebung erfolgte über ein haptisches Eingabegerät. Während der Simulation wurden die Parameter Eingriffsdauer und Fehleranzahl erfasst. Zehn Probanden absolvierten jeweils eine Trainingseinheit bestehend aus je 5 Übungsdurchläufen an 10 aufeinanderfolgenden Tagen.
Ergebnisse: Vier Probanden verringerten die benötigte Zeit um mehr als 60% im Verlauf des Übungszeitraums. Vier der Probanden verringerten ihre Fehleranzahl um mehr als 60%. Acht von 10 Probanden zeigten eine Verbesserung bezüglich beider Parameter. Im Median wurde im gesamten gemessenen Zeitraum die Dauer des Eingriffs um 46 Sekunden und die Fehleranzahl um 191 reduziert. Die Überprüfung eines Zusammenhangs zwischen den 2 Parametern ergab eine positive Korrelation.
Schlussfolgerung: Zusammenfassend lässt sich feststellen, dass das Training am NNH-Simulator auch bei unerfahrenen Personen die Performance beträchtlich verbessert, sowohl in Bezug auf die Dauer als auch auf die Genauigkeit des Eingriffs.
Zur Unterstützung des Operateurs wird eine patientennahe Informationsanzeige entwickelt, die kontextrelevante Informationen entsprechend der aktuellen Situation bereitstellen kann. Hierfür soll eine Situationserkennung konzipiert werden, die auf unterschiedliche intraoperative Prozesse übertragen werden kann. Ziel der adaptiven Situationserkennung ist das Erkennen spezifischer Situationen durch intraoperative Informationen unterschiedlicher Datenquellen im Operationssaal. Innerhalb der Datenerhebung und -analyse wurden Anwendungsfälle für die Situationserkennung definiert sowie chirurgische Prozessmodelle erstellt, die intraoperative Ereignisse abbilden. Auf Basis dieser Informationen wurde ein Konzept entworfen, das sich zunächst auf die Erkennung abstrakter generalisierter Phasen, unabhängig vom Eingriff, fokussiert und sich Schritt für Schritt auf granulare Prozessschritte spezifizieren lässt. Diese Flexibilität soll die Übertragbarkeit des Konzepts auf intraoperative Prozesse ermöglichen und den Operateur dadurch gezielt mit kontextrelevanten Informationen unterstützen. Das Konzept wird in zukünftigen Schritten weiterentwickelt.
Glioblastomas are the most aggressive fast-growing primary brain cancer which originate in the glial cells of the brain. Accurate identification of the malignant brain tumor and its sub-regions is still one of the most challenging problems in medical image segmentation. The Brain Tumor Segmentation Challenge (BraTS) has been a popular benchmark for automatic brain glioblastomas segmentation algorithms since its initiation. In this year, BraTS 2021 challenge provides the largest multi-parametric (mpMRI) dataset of 2,000 pre-operative patients. In this paper, we propose a new aggregation of two deep learning frameworksnamely, DeepSeg and nnU-Net for automatic glioblastoma recognition in pre-operative mpMRI. Our ensemble method obtains Dice similarity scores of 92.00, 87.33, and 84.10 and Hausdorff Distances of 3.81, 8.91, and 16.02 for the enhancing tumor, tumor core, and whole tumor regions, respectively, on the BraTS 2021 validation set, ranking us among the top ten teams. These experimental findings provide evidence that it can be readily applied clinically and thereby aiding in the brain cancer prognosis, therapy planning, and therapy response monitoring. A docker image for reproducing our segmentation results is available online at (https://hub.docker.com/r/razeineldin/deepseg21).
Recent advances in artificial intelligence have enabled promising applications in neurosurgery that can enhance patient outcomes and minimize risks. This paper presents a novel system that utilizes AI to aid neurosurgeons in precisely identifying and localizing brain tumors. The system was trained on a dataset of brain MRI scans and utilized deep learning algorithms for segmentation and classification. Evaluation of the system on a separate set of brain MRI scans demonstrated an average Dice similarity coefficient of 0.87. The system was also evaluated through a user experience test involving the Department of Neurosurgery at the University Hospital Ulm, with results showing significant improvements in accuracy, efficiency, and reduced cognitive load and stress levels. Additionally, the system has demonstrated adaptability to various surgical scenarios and provides personalized guidance to users. These findings indicate the potential for AI to enhance the quality of neurosurgical interventions and improve patient outcomes. Future work will explore integrating this system with robotic surgical tools for minimally invasive surgeries.
The focus of the developed maturity model was set on processes. The concept of the widespread CMM and its practices has been transferred to the perioperative domain and the concept of the new maturity model. Additional optimization goals and technological as well as networking-specific aspects enable a process- and object-focused view of the maturity model in order to ensure broad coverage of different subareas. The evaluation showed that the model is applicable to the perioperative field. Adjustments and extensions of the maturity model are future steps to improve the rating and classification of the new maturity model.
Uncontrolled movements of laparoscopic instruments can lead to inadvertent injury of adjacent structures. The risk becomes evident when the dissecting instrument is located outside the field of view of the laparoscopic camera. Technical solutions to ensure patient safety are appreciated. The present work evaluated the feasibility of an automated binary classification of laparoscopic image data using Convolutional Neural Networks (CNN) to determine whether the dissecting instrument is located within the laparoscopic image section. A unique record of images was generated from six laparoscopic cholecystectomies in a surgical training environment to configure and train The CNN. By using a temporary version of the neural network, the annotation of the training image files could be automated and accelerated. A combination of oversampling and selective data augmentation was used to enlarge the fully labelled image data set and prevent loss of accuracy due to imbalanced class volumes. Subsequently the same approach was applied to the comprehensive, fully annotated Cholec80 database. The described process led to the generation of extensive and balanced training image data sets. The performance of the CNN-based binary classifiers was evaluated on separate test records from both databases. On our recorded data, an accuracy of 0.88 with regard to the safety-relevant classification was achieved. The subsequent evaluation on the Cholec80 data set yielded an accuracy of 0.84. The presented results demonstrate the feasibility of a binary classification of laparoscopic image data for the detection of adverse events in a surgical training environment using a specifically configured CNN architecture.
Clinical reading centers provide expertise for consistent, centralized analysis of medical data gathered in a distributed context. Accordingly, appropriate software solutions are required for the involved communication and data management processes. In this work, an analysis of general requirements and essential architectural and software design considerations for reading center information systems is provided. The identified patterns have been applied to the implementation of the reading center platform which is currently operated at the Center of Ophthalmology of the University Hospital of Tübingen.