Explainable Artificial Intelligence for Deep-Learning Based Classification of Cystic Fibrosis Lung Changes in MRI.

Algorithms increasing the transparence and explain ability of neural networks are gaining more popularity. Applying them to custom neural network architectures and complex medical problems remains challenging. In this work, several algorithms such as integrated gradients and grad came were used to generate additional explainable outputs for the classification of lung perfusion changes and mucus plugging in cystic fibrosis patients on MRI. The algorithms are applied on top of an already existing deep learning-based classification pipeline. From six explain ability algorithms, four were implemented successfully and one yielded satisfactory results which might provide support to the radiologist. It was evident, that the areas relevant for the classification were highlighted, thus emphasizing the applicability of deep learning for classification of lung changes in CF patients. Using explainable concepts with deep learning could improve confidence of clinicians towards deep learning and introduction of more diagnostic decision support systems.

Next-generation study databases require FAIR, EHR-integrated, and scalable Electronic Data Capture for medical documentation and decision support.

Structured patient data play a key role in all types of clinical research. They are often collected in study databases for research purposes. In order to describe characteristics of a next-generation study database and assess the feasibility of its implementation a proof-of-concept study in a German university hospital was performed. Key characteristics identified include FAIR access to electronic case report forms (eCRF), regulatory compliant Electronic Data Capture (EDC), an EDC with electronic health record (EHR) integration, scalable EDC for medical documentation, patient generated data, and clinical decision support. In a local case study, we then successfully implemented a next-generation study database for 19 EDC systems (n = 2217 patients) that linked to i.s.h.med (Oracle Cerner) with the local EDC system called OpenEDC. Desiderata of next-generation study databases for patient data were identified from ongoing local clinical study projects in 11 clinical departments at Heidelberg University Hospital, Germany, a major tertiary referral hospital. We compiled and analyzed feature and functionality requests submitted to the OpenEDC team between May 2021 and July 2023. Next-generation study databases are technically and clinically feasible. Further research is needed to evaluate if our approach is feasible in a multi-center setting as well.

Automated calculation of ontology-based planning proposals: An application in reconstructive oral and maxillofacial surgery.

BACKGROUND: Structured modelling of surgical knowledge and its automated processing is still challenging. The aim of this work is to introduce a novel approach for automated calculation of ontology-based planning proposals in mandibular reconstruction and conduct a feasibility study. METHODS: The presented approach is composed of an RDF(S) ontology, a 3D mandible template and a calculator-optimiser algorithm to automatically calculate reconstruction proposals with fibula grafts. To validate the viability of the approach, a feasibility study was conducted on 164 simulated mandibular reconstructions. RESULTS: The ontology defines 244 different reconstruction variants and 80 analyses for optimization. In 146 simulated cases, a proposal could be automatically calculated (average time 8.79 ± 4.03 s). The assessments of the proposals by three clinical experts indicate the viability of the approach. CONCLUSIONS: Due to the modular separation between computational logic and domain knowledge, the developed concepts can be easily maintained, reused and adapted for other applications.

3D-2D Distance Maps Conversion Enhances Classification of Craniosynostosis.

OBJECTIVE: Diagnosis of craniosynostosis using photogrammetric 3D surface scans is a promising radiation-free alternative to traditional computed tomography. We propose a 3D surface scan to 2D distance map conversion enabling the usage of the first convolutional neural networks (CNNs)-based classification of craniosynostosis. Benefits of using 2D images include preserving patient anonymity, enabling data augmentation during training, and a strong under-sampling of the 3D surface with good classification performance. METHODS: The proposed distance maps sample 2D images from 3D surface scans using a coordinate transformation, ray casting, and distance extraction. We introduce a CNN-based classification pipeline and compare our classifier to alternative approaches on a dataset of 496 patients. We investigate into low-resolution sampling, data augmentation, and attribution mapping. RESULTS: Resnet18 outperformed alternative classifiers on our dataset with an F1-score of 0.964 and an accuracy of 98.4%. Data augmentation on 2D distance maps increased performance for all classifiers. Under-sampling allowed 256-fold computation reduction during ray casting while retaining an F1-score of 0.92. Attribution maps showed high amplitudes on the frontal head. CONCLUSION: We demonstrated a versatile mapping approach to extract a 2D distance map from the 3D head geometry increasing classification performance, enabling data augmentation during training on 2D distance maps, and the usage of CNNs. We found that low-resolution images were sufficient for a good classification performance. SIGNIFICANCE: Photogrammetric surface scans are a suitable craniosynostosis diagnosis tool for clinical practice. Domain transfer to computed tomography seems likely and can further contribute to reducing ionizing radiation exposure for infants.

The Use of Artificial Intelligence for the Classification of Craniofacial Deformities.

Positional cranial deformities are a common finding in toddlers, yet differentiation from craniosynostosis can be challenging. The aim of this study was to train convolutional neural networks (CNNs) to classify craniofacial deformities based on 2D images generated using photogrammetry as a radiation-free imaging technique. A total of 487 patients with photogrammetry scans were included in this retrospective cohort study: children with craniosynostosis (n = 227), positional deformities (n = 206), and healthy children (n = 54). Three two-dimensional images were extracted from each photogrammetry scan. The datasets were divided into training, validation, and test sets. During the training, fine-tuned ResNet-152s were utilized. The performance was quantified using tenfold cross-validation. For the detection of craniosynostosis, sensitivity was at 0.94 with a specificity of 0.85. Regarding the differentiation of the five existing classes (trigonocephaly, scaphocephaly, positional plagiocephaly left, positional plagiocephaly right, and healthy), sensitivity ranged from 0.45 (positional plagiocephaly left) to 0.95 (scaphocephaly) and specificity ranged from 0.87 (positional plagiocephaly right) to 0.97 (scaphocephaly). We present a CNN-based approach to classify craniofacial deformities on two-dimensional images with promising results. A larger dataset would be required to identify rarer forms of craniosynostosis as well. The chosen 2D approach enables future applications for digital cameras or smartphones.

Impact of data synthesis strategies for the classification of craniosynostosis.

Introduction: Photogrammetric surface scans provide a radiation-free option to assess and classify craniosynostosis. Due to the low prevalence of craniosynostosis and high patient restrictions, clinical data are rare. Synthetic data could support or even replace clinical data for the classification of craniosynostosis, but this has never been studied systematically. Methods: We tested the combinations of three different synthetic data sources: a statistical shape model (SSM), a generative adversarial network (GAN), and image-based principal component analysis for a convolutional neural network (CNN)-based classification of craniosynostosis. The CNN is trained only on synthetic data but is validated and tested on clinical data. Results: The combination of an SSM and a GAN achieved an accuracy of 0.960 and an F1 score of 0.928 on the unseen test set. The difference to training on clinical data was smaller than 0.01. Including a second image modality improved classification performance for all data sources. Conclusions: Without a single clinical training sample, a CNN was able to classify head deformities with similar accuracy as if it was trained on clinical data. Using multiple data sources was key for a good classification based on synthetic data alone. Synthetic data might play an important future role in the assessment of craniosynostosis.

A Radiation-Free Classification Pipeline for Craniosynostosis Using Statistical Shape Modeling.

BACKGROUND: Craniosynostosis is a condition caused by the premature fusion of skull sutures, leading to irregular growth patterns of the head. Three-dimensional photogrammetry is a radiation-free alternative to the diagnosis using computed tomography. While statistical shape models have been proposed to quantify head shape, no shape-model-based classification approach has been presented yet. METHODS: We present a classification pipeline that enables an automated diagnosis of three types of craniosynostosis. The pipeline is based on a statistical shape model built from photogrammetric surface scans. We made the model and pathology-specific submodels publicly available, making it the first publicly available craniosynostosis-related head model, as well as the first focusing on infants younger than 1.5 years. To the best of our knowledge, we performed the largest classification study for craniosynostosis to date. RESULTS: Our classification approach yields an accuracy of 97.8 %, comparable to other state-of-the-art methods using both computed tomography scans and stereophotogrammetry. Regarding the statistical shape model, we demonstrate that our model performs similar to other statistical shape models of the human head. CONCLUSION: We present a state-of-the-art shape-model-based classification approach for a radiation-free diagnosis of craniosynostosis. Our publicly available shape model enables the assessment of craniosynostosis on realistic and synthetic data.

CNN-Based Classification of Craniosynostosis Using 2D Distance Maps.

Craniosynostosis is a condition associated with the premature fusion of skull sutures affecting infants. 3D photogrammetric scans are a promising alternative to computed tomography scans in cases of single suture or nonsyndromic synostosis for diagnostic imaging, but oftentimes diagnosis is not automated and relies on additional cephalometric measure-ments and the experience of the surgeon. We propose an alternative representation of the infant's head shape created from 3D photogrammetric surface scans as 2D distance maps. Those 2D distance maps rely on ray casting to extract distances from a center point to the head surface, arranging them into a 2D image grid. We use the distance map for an original convolutional neural network (CNN)-based classification approach, which is evaluated on a publicly available synthetic dataset for benchmarking and also tested on clinical data. Qualitative differences of different head shapes can be ob-served in the distance maps. The CNN-based classifier achieves accuracies of 100 % on the publicly available synthetic dataset and 98.86 % on the clinical test set. Our distance map approach demonstrates the diagnostic value of 3D photogrammetry and the possibility of automatic, CNN-based diagnosis. Future steps include the improvement of the mapping method and testing the CNN on more pathologies.

Knowledge Acquisition and Construction of a RDF-Ontology for Computer-Assisted Surgery.

The integration of surgical knowledge into virtual planning systems plays a key role in computer-assisted surgery. The knowledge is often implicitly contained in the implemented algorithms. However, a strict separation would be desirable for reasons of maintainability, reusability and readability. Along with the Department of Oral and Maxillofacial Surgery at Heidelberg University Hospital, we are working on the development of a virtual planning system for mandibular reconstruction. In this work we describe a process for the structured acquisition and representation of surgical knowledge for mandibular reconstruction. Based on the acquired knowledge, an RDF(S) ontology was created. The ontology is connected to the virtual planning system via a SPARQL interface. The described process of knowledge acquisition can be transferred to other surgical use cases. Furthermore, the developed ontology is characterised by a reusable and easily expandable data model.

eHealth and Clinical Documentation Systems.

eHealth is the use of modern information and communication technology (ICT) for trans-institutional healthcare purposes. Important subtopics of eHealth are health data sharing and telemedicine. Most of the clinical documentation to be shared is collected in patient records to support patient care. More sophisticated approaches to electronic patient records are trans-institutional or (inter-)national. Other aims for clinical documentation are quality management, reimbursement, legal issues, and medical research. Basic prerequisite for eHealth is interoperability, which can be divided into technical, semantic and process interoperability. There is a variety of international standards to support interoperability. Telemedicine is a subtopic of eHealth, which bridges spatial distance by using ICT for medical (inter-)actions. We distinguish telemedicine among healthcare professionals and telemedicine between health care professionals and patients. Both have a great potential to face the challenges of aging societies, the increasing number of chronically ill patients, multimorbidity and low number of physicians in remote areas. With ongoing digitalization more and more data are available digitally. Clinical documentation is an important source for big data analysis and artificial intelligence. The patient has an important role: Telemonitoring, wearable technologies, and smart home devices provide digital health data from daily life. These are high-quality data which can be used for medical decisions.

On the Construction of Multilingual Corpora for Clinical Text Mining.

The amount of digital data derived from healthcare processes have increased tremendously in the last years. This applies especially to unstructured data, which are often hard to analyze due to the lack of available tools to process and extract information. Natural language processing is often used in medicine, but the majority of tools used by researchers are developed primarily for the English language. For developing and testing natural language processing methods, it is important to have a suitable corpus, specific to the medical domain that covers the intended target language. To improve the potential of natural language processing research, we developed tools to derive language specific medical corpora from publicly available text sources. n order to extract medicine-specific unstructured text data, openly available pub-lications from biomedical journals were used in a four-step process: (1) medical journal databases were scraped to download the articles, (2) the articles were parsed and consolidated into a single repository, (3) the content of the repository was de-scribed, and (4) the text data and the codes were released. In total, 93 969 articles were retrieved, with a word count of 83 868 501 in three different languages (German, English, and Spanish) from two medical journal databases Our results show that unstructured text data extraction from openly available medical journal databases for the construction of unified corpora of medical text data can be achieved through web scraping techniques.

Electromagnetic navigated condylar positioning after high oblique sagittal split osteotomy of the mandible: a guided method to attain pristine temporomandibular joint conditions.

OBJECTIVES: Reproduction of the exact preoperative proximal-mandible position after osteotomy in orthognathic surgery is difficult to achieve. This clinical pilot study evaluated an electromagnetic (EM) navigation system for condylar positioning after high-oblique sagittal split osteotomy (HSSO). STUDY DESIGN: After HSSO as part of 2-jaw surgery, the position of 10 condyles was intraoperatively guided by an EM navigation system. As controls, 10 proximal segments were positioned by standard manual replacement. Accuracy was measured by pre- and postoperative cone beam computed tomography imaging. RESULTS: Overall, EM condyle repositioning was equally accurate compared with manual repositioning (P > .05). Subdivided into 3 axes, significant differences could be identified (P < .05). Nevertheless, no significantly and clinically relevant dislocations of the proximal segment of either the EM or the manual repositioning method could be shown (P > .05). CONCLUSIONS: This pilot study introduces a guided method for proximal segment positioning after HSSO by applying the intraoperative EM system. The data demonstrate the high accuracy of EM navigation, although manual replacement of the condyles could not be surpassed. However, EM navigation can avoid clinically hidden, severe malpositioning of the condyles.

Computer assisted positioning of the proximal segment after sagittal split osteotomy of the mandible: Preclinical investigation of a novel electromagnetic navigation system.

INTRODUCTION: Modifications of the temporomandibular joint position after mandible osteotomy are reluctantly accepted in orthognathic surgery. To tackle this problem, we developed a new navigation system using miniaturized electromagnetic sensors. Our imageless navigation approach is therefore optimized to avoid complications of previously proposed optical approaches such as the interference with established surgical procedures and the line of sight problem. MATERIAL AND METHODS: High oblique sagittal split osteotomies were performed on 6 plastic skull mandibles in a laboratory under conditions comparable to the operating theatre. The subsequent condyle reposition was guided by an intuitive user interface and performed by electromagnetic navigation. To prove the suitability and accuracy of this novel approach for condyle navigation, the positions of 3 titanium marker screws placed on each of the proximal segments were compared using pre- and postoperative Cone Beam Computed Tomography (CBCT) imaging. RESULTS: Guided by the electromagnetic navigation system, positioning of the condyles was highly accurate in all dimensions. Translational discrepancies up to 0,65 mm and rotations up to 0,38° in mean could be measured postoperatively. There were no statistically significant differences between navigation results and CBCT measurements. CONCLUSION: The intuitive user interface provides a simple way to precisely restore the initial position and orientation of the proximal mandibular segments. Our electromagnetic navigation system therefore yields a promising approach for orthognathic surgery applications.

Electromagnetic navigated positioning of the maxilla after Le Fort I osteotomy in preclinical orthognathic surgery cases.

OBJECTIVES: Inaccuracies in orthognathic surgery can be caused during face-bow registration, model surgery on plaster models, and intermaxillary splint manufacturing. Electromagnetic (EM) navigation is a promising method for splintless digitized maxillary positioning. STUDY DESIGN: After performing Le Fort I osteotomy on 10 plastic skulls, the target position of the maxilla was guided by an EM navigation system. Specially implemented software illustrated the target position by real-time multistage colored three-dimensional imaging. Accuracy was determined by using pre- and postoperative cone beam computed tomography. RESULTS: The high accuracy of the EM system was underlined by the fact that it had a navigated maxilla position discrepancy of only 0.4 mm, which was verified by postoperative cone beam computed tomography. CONCLUSIONS: This preclinical study demonstrates a precise digitized approach for splintless maxillary repositioning after Le Fort I osteotomy. The accuracy and intuitive illustration of the introduced EM navigation system is promising for potential daily use in orthognathic surgery.

Can electromagnetic-navigated maxillary positioning replace occlusional splints in orthognathic surgery? A clinical pilot study.

INTRODUCTION: Because of the inaccuracy of intermaxillary splints in orthognathic surgery, intraoperative guidance via a real time navigation system might represent a suitable method for enhancing the precision of maxillary positioning. Therefore, in this clinical trial, maxillary repositioning after Le Fort I osteotomy was guided splintless by an electromagnetic navigation system. MATERIALS AND METHODS: Conservatively planned maxillary reposition in each of 5 patients was transferred to a novel software module of the electromagnetic navigation system. Intraoperatively, after Le Fort I osteotomy, the software guided the maxilla to the targeted position. Accuracy was evaluated by pre- and postoperative cone beam computer tomography imaging (the vectorial distance of the incisal marker points was measured in three dimensions) and compared with that of a splint transposed control group. RESULTS: The repositioning of the maxilla guided by the electromagnetic navigation system was intuitive and simple to accomplish. The achieved maxillary position with a deviation of 0.7 mm on average to the planned position was equally accurate compared with that of the splint transposed control group of 0.5 mm (p > 0.05). DISCUSSION: The data of this clinical study display good accuracy for splintless electromagnetic-navigated maxillary positioning. Nevertheless, this method does not surpass the splint-encoded gold standard with regard to accuracy. Future investigations will be necessary to show the full potential of electromagnetic navigation in orthognathic surgery.

Pre- and intraoperative processing and integration of various anatomical and functional data in neurosurgery.

A software system is presented, capable of integrating various information sources for neurosurgical procedures. These include anatomical data such as a standard 3D DICOM image stacks, atlas data, as well as functional information (e.g. fmri, MEG, EEG). The system is programmed in C++ using Open GL for visualisation, and was developed in a close cooperation with a neurosurgical department to match existing needs. Preoperatively the data may be combined, registered by a rigid or elastic matching process, enriched by user specified planning information such as annotations or trajectories, and visualised in a standard fashion using different segmentation schemes, interactive rotation, zooming etc. Selected portions of the gathered and generated information may then be exported for neuronavigation input in DICOM or a vendor specific format. Intraoperatively, this information may, on the one hand, be simply used as an integrational part of the routinely used navigational data. On the other hand, the system is also capable of interacting with the navigational system to integrate the actual spatial information of the ongoing procedure into the preoperative data, thus allowing further planning and visualisation beyond the scope of the navigational unit. Furthermore, intraoperatively updated information such as intraoperative MR images or electrophysiological data may be integrated and correlated to the existing information. Ongoing developments comprise redistribution of the relevant data for injection onto the screen of the navigational system or into the optical pathway of the (3D capable) microscope display/ocular. This also includes information about the actual automatically optimized accuracy of the navigation process in relation to the head markers in use.

Approach to intraoperative electromagnetic navigation in orthognathic surgery: A phantom skull based trial.

INTRODUCTION: Intraoperative guidance using electromagnetic navigation is an upcoming method in maxillofacial surgery. However, due to their unwieldy structures, especially the line-of-sight problem, optical navigation devices are not used for daily orthognathic surgery. Therefore, orthognathic surgery was simulated on study phantom skulls, evaluating the accuracy and handling of a new electromagnetic tracking system. MATERIAL AND METHODS: Le-Fort I osteotomies were performed on 10 plastic skulls. Orthognathic surgical planning was done in the conventional way using plaster models. Accuracy of the gold standard, splint-based model surgery versus an electromagnetic tracking system was evaluated by measuring the actual maxillary deviation using bimaxillary splints and preoperative and postoperative cone beam computer tomography imaging. The distance of five anatomical marker points were compared pre- and postoperatively. RESULTS: The electromagnetic tracking system was significantly more accurate in all measured parameters compared with the gold standard using bimaxillary splints (p < 0.01). The data shows a discrepancy between the model surgical plans and the actual correction of the upper jaw of 0.8 mm. Using the electromagnetic tracking, we could reduce the discrepancy of the maxillary transposition between the planned and actual orthognathic surgery to 0.3 mm on average. DISCUSSION: The data of this preliminary study shows a high level of accuracy in surgical orthognathic performance using electromagnetic navigation, and may offer greater precision than the conventional plaster model surgery with bimaxillary splints. CONCLUSION: This preliminary work shows great potential for the establishment of an intraoperative electromagnetic navigation system for maxillofacial surgery.

Semantic focusing allows fully automated single-layer slide scanning of cervical cytology slides.

Liquid-based cytology (LBC) in conjunction with Whole-Slide Imaging (WSI) enables the objective and sensitive and quantitative evaluation of biomarkers in cytology. However, the complex three-dimensional distribution of cells on LBC slides requires manual focusing, long scanning-times, and multi-layer scanning. Here, we present a solution that overcomes these limitations in two steps: first, we make sure that focus points are only set on cells. Secondly, we check the total slide focus quality. From a first analysis we detected that superficial dust can be separated from the cell layer (thin layer of cells on the glass slide) itself. Then we analyzed 2,295 individual focus points from 51 LBC slides stained for p16 and Ki67. Using the number of edges in a focus point image, specific color values and size-inclusion filters, focus points detecting cells could be distinguished from focus points on artifacts (accuracy 98.6%). Sharpness as total focus quality of a virtual LBC slide is computed from 5 sharpness features. We trained a multi-parameter SVM classifier on 1,600 images. On an independent validation set of 3,232 cell images we achieved an accuracy of 94.8% for classifying images as focused. Our results show that single-layer scanning of LBC slides is possible and how it can be achieved. We assembled focus point analysis and sharpness classification into a fully automatic, iterative workflow, free of user intervention, which performs repetitive slide scanning as necessary. On 400 LBC slides we achieved a scanning-time of 13.9±10.1 min with 29.1±15.5 focus points. In summary, the integration of semantic focus information into whole-slide imaging allows automatic high-quality imaging of LBC slides and subsequent biomarker analysis.