Performance evaluation of three versions of a convolutional neural network for object detection and segmentation using a multiclass and reduced panoramic radiograph dataset.

OBJECTIVES: To evaluate the diagnostic performance of three versions of a deep-learning convolutional neural network in terms of object detection and segmentation using a multiclass panoramic radiograph dataset. METHODS: A total of 600 orthopantomographies were randomly selected for this study and manually annotated by a single operator using an image annotation tool (COCO Annotator v.11.0.1) to establish ground truth. The annotation classes included teeth, maxilla, mandible, inferior alveolar nerve, dento- and implant-supported crowns/pontics, endodontic treatment, resin-based restorations, metallic restorations, and implants. The dataset was then divided into training, validation, and testing subsets, which were used to train versions 5, 7, and 8 of You Only Look Once (YOLO) Neural Network. Results were stored, and a posterior performance analysis was carried out by calculating the precision (P), recall (R), F1 Score, Intersection over Union (IoU), and mean average precision (mAP) at 0.5 and 0.5-0.95 thresholds. The confusion matrix and recall precision graphs were also sketched. RESULTS: YOLOv5s showed an improvement in object detection results with an average R = 0.634, P = 0.781, mAP0.5 = 0.631, and mAP0.5-0.95 = 0.392. YOLOv7m achieved the best object detection results with average R = 0.793, P = 0.779, mAP0.5 = 0.740, and mAP0.5-0.95 = 0,481. For object segmentation, YOLOv8m obtained the best average results (R = 0.589, P = 0.755, mAP0.5 = 0.591, and mAP0.5-0.95 = 0.272). CONCLUSIONS: YOLOv7m was better suited for object detection, while YOLOv8m demonstrated superior performance in object segmentation. The most frequent error in object detection was related to background classification. Conversely, in object segmentation, there is a tendency to misclassify True Positives across different dental treatment categories. CLINICAL SIGNIFICANCE: General diagnostic and treatment decisions based on panoramic radiographs can be enhanced using new artificial intelligence-based tools. Nevertheless, the reliability of these neural networks should be subjected to training and validation to ensure their generalizability.

SNOMED2HL7: A tool to normalize and bind SNOMED CT concepts to the HL7 Reference Information Model.

BACKGROUND: Current clinical research and practice requires interoperability among systems in a complex and highly dynamic domain. There has been a significant effort in recent years to develop integrative common data models and domain terminologies. Such efforts have not completely solved the challenges associated with clinical data that are distributed among different and heterogeneous institutions with different systems to encode the information. Currently, when providing homogeneous interfaces to exploit clinical data, certain transformations still involve manual and time-consuming processes that could be automated. OBJECTIVES: There is a lack of tools to support data experts adopting clinical standards. This absence is especially significant when links between data model and vocabulary are required. The objective of this work is to present SNOMED2HL7, a novel tool to automatically link biomedical concepts from widely used terminologies, and the corresponding clinical context, to the HL7 Reference Information Model (RIM). METHODS: Based on the recommendations of the International Health Terminology Standards Development Organisation (IHTSDO), the SNOMED Normal Form has been implemented within SNOMED2HL7 to decompose and provide a method to reduce the number of options to store the same information. The binding of clinical terminologies to HL7 RIM components is the core of SNOMED2HL7, where terminology concepts have been annotated with the corresponding options within the interoperability standard. A web-based tool has been developed to automatically provide information from the normalization mechanisms and the terminology binding. RESULTS: SNOMED2HL7 binding coverage includes the majority of the concepts used to annotate legacy systems. It follows HL7 recommendations to solve binding overlaps and provides the binding of the normalized version of the concepts. The first version of the tool, available at http://kandel.dia.fi.upm.es:8078, has been validated in EU funded projects to integrate real world data for clinical research with an 88.47% of accuracy. CONCLUSIONS: This paper presents the first initiative to automatically retrieve concept-centered information required to transform legacy data into widely adopted interoperability standards. Although additional functionality will extend capabilities to automate data transformations, SNOMED2HL7 already provides the functionality required for the clinical interoperability community.