In radiation therapy, fiducial markers implanted in a patient’s body are tracked using X-ray fluoroscopy to estimate tumor positions. However, flexible markers, such as Gold Anchor®(Naslund Medical AB, Sweden), deform within the body, making conventional template matching challenging. While deep learning offers a promising solution, the extensive collection and annotation of clinical data required for training pose a significant barrier to adoption. To address this, we propose a tracking framework that utilizes a lightweight Siamese CNN trained exclusively on synthetic fluoroscopy images. Our method generates synthetic data simulating diverse marker deformations under low-contrast and high-noise conditions, employs dynamic programming for stable initial detection, and performs real-time tracking with a particle filter. In evaluations using clinical data, our method achieves a tracking accuracy of 0.42 ± 0.12 pixels for prostate cancer cases and 0.97 ± 0.53 pixels for pancreatic cancer cases. This significantly outperforms conventional methods, particularly in challenging low-contrast pancreatic cancer cases. With TensorRT optimization, the framework achieves a processing speed of 3.8 ms/frame. This work presents a practical solution for high-accuracy tracking, reducing data collection costs and facilitating the use of deep learning in clinical applications.

SUMMARY: : Similarity searches of amino acid sequences against the public metagenomic data can provide users insights about the function of sequences based on the environmental distribution of similar sequences. However, a considerable reduction in the amount of data or the accuracy of the result was necessary to conduct sequence similarity searches against public metagenomic data, because of the vast data size more than Terabytes. Here, we present an ultra-fast service for the highly accurate amino acid sequence similarity search, called PZLAST, which can search the user's amino acid sequences to several Terabytes of public metagenomic sequences in ∼10-20 min. PZLAST accomplishes its search speed by using PEZY-SC2, which is a Multiple Instruction Multiple Data many-core processor. Results of PZLAST are summarized by the ontology-based environmental distribution of similar sequences. PZLAST can be used to predict the function of sequences and mine for homologs of functionally important gene sequences. AVAILABILITY AND IMPLEMENTATION: PZLAST is freely accessible at https://pzlast.riken.jp/meta without requiring registration. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.