METHODS: This bibliometric work investigated the academic publication trends in medical image segmentation technology. These data were collected from the Web of Science (WoS) Core Collection and the Scopus. In the quantitative analysis stage, important visual maps were produced to show publication trends from five different perspectives including annual publications, countries, top authors, publication sources, and keywords. In the qualitative analysis stage, the frequently used methods and research trends in the medical image segmentation field were analyzed from 49 publications with the top annual citation rates.
RESULTS: The analysis results showed that the number of publications had increased rapidly by year. The top related countries include the Chinese mainland, the United States, and India. Most of these publications were conference papers, besides there are also some top journals. The research hotspot in this field was deep learning-based medical image segmentation algorithms based on keyword analysis. These publications were divided into three categories: reviews, segmentation algorithm publications, and other relevant publications. Among these three categories, segmentation algorithm publications occupied the vast majority, and deep learning neural network-based algorithm was the research hotspots and frontiers.
CONCLUSIONS: Through this bibliometric research work, the research hotspot in the medical image segmentation field is uncovered and can point to future research in the field. It can be expected that more researchers will focus their work on deep learning neural network-based medical image segmentation.
METHODS: In our paper, we propose a real-time, lightweight liver segmentation model named G-MBRMD. Specifically, we employ a Transformer-based complex model as the teacher and a convolution-based lightweight model as the student. By introducing proposed multi-head mapping and boundary reconstruction strategies during the knowledge distillation process, Our method effectively guides the student model to gradually comprehend and master the global boundary processing capabilities of the complex teacher model, significantly enhancing the student model's segmentation performance without adding any computational complexity.
RESULTS: On the LITS dataset, we conducted rigorous comparative and ablation experiments, four key metrics were used for evaluation, including model size, inference speed, Dice coefficient, and HD95. Compared to other methods, our proposed model achieved an average Dice coefficient of 90.14±16.78%, with only 0.6 MB memory and 0.095 s inference speed for a single image on a standard CPU. Importantly, this approach improved the average Dice coefficient of the baseline student model by 1.64% without increasing computational complexity.
CONCLUSION: The results demonstrate that our method successfully realizes the unification of segmentation precision and lightness, and greatly enhances its potential for widespread application in practical settings.
SYSTEMATIC REVIEW REGISTRATION: https://www.crd.york.ac.uk/prospero/, identifier CRD42024559872.