RESUMO
Domain adaptation is a subfield of statistical learning theory that takes into account the shift between the distribution of training and test data, typically known as source and target domains, respectively. In this context, this paper presents an incremental approach to tackle the intricate challenge of unsupervised domain adaptation, where labeled data within the target domain is unavailable. The proposed approach, OTP-DA, endeavors to learn a sequence of joint subspaces from both the source and target domains using Linear Discriminant Analysis (LDA), such that the projected data into these subspaces are domain-invariant and well-separated. Nonetheless, the necessity of labeled data for LDA to derive the projection matrix presents a substantial impediment, given the absence of labels within the target domain in the setting of unsupervised domain adaptation. To circumvent this limitation, we introduce a selective label propagation technique grounded on optimal transport (OTP), to generate pseudo-labels for target data, which serve as surrogates for the unknown labels. We anticipate that the process of inferring labels for target data will be substantially streamlined within the acquired latent subspaces, thereby facilitating a self-training mechanism. Furthermore, our paper provides a rigorous theoretical analysis of OTP-DA, underpinned by the concept of weak domain adaptation learners, thereby elucidating the requisite conditions for the proposed approach to solve the problem of unsupervised domain adaptation efficiently. Experimentation across a spectrum of visual domain adaptation problems suggests that OTP-DA exhibits promising efficacy and robustness, positioning it favorably compared to several state-of-the-art methods.
Assuntos
Aprendizado de Máquina não Supervisionado , Humanos , Análise Discriminante , AlgoritmosRESUMO
INTRODUCTION: Wrist arthroscopy is a rapidly expanding surgical discipline, but has a long and challenging learning curve. One of its difficulties is distinguishing the various anatomical structures during the procedure. Although artificial intelligence has made significant progress in recent decades, its potential as a valuable tool in surgery training is largely untapped. MATERIALS AND METHODS: The objective of this study was to develop an algorithm that could accurately recognize the anatomical bone structures of the wrist during arthroscopy. We prospectively included 20 wrist arthroscopies: 10 in patients and 10 in cadavers. For each surgery, we extracted and labeled images of the various carpal bones. These images were used to create a database for training, validating and testing a structure recognition algorithm. The primary criterion used was a Dice loss detection and categorization score for structures of interest, with a threshold greater than 80%. RESULTS: The database contained 511 labeled images (4,088 after data augmentation). We developed a Deeplabv3+ classification algorithm with a U-Net architecture. After training and testing our algorithm, we achieved an average Dice loss score of 89% for carpal bone recognition. CONCLUSION: This study demonstrated reliable detection of different carpal bones during arthroscopic wrist surgery using artificial intelligence. However, some bones were detected more accurately than others, suggesting that additional algorithm training could further enhance performance. Application in real-life conditions could validate these results and potentially contribute to learning and improvement in arthroscopic wrist surgery. LEVEL OF EVIDENCE: IV.
RESUMO
The exponential growth of data generates terabytes of very large databases. The growing number of data dimensions and data objects presents tremendous challenges for effective data analysis and data exploration methods and tools. Thus, it becomes crucial to have methods able to construct a condensed description of the properties and structure of data, as well as visualization tools capable of representing the data structure from these condensed descriptions. The purpose of our work described in this paper is to develop a method of describing data from enriched and segmented prototypes using a topological clustering algorithm. We then introduce a visualization tool that can enhance the structure within and between groups in data. We show, using some artificial and real databases, the relevance of the proposed approach.