A New Registration Approach for Dynamic Analysis of Calcium Signals in Organs
Wing pouches of fruit flies are a powerful genetic model system that has been used widely for studying the conserved biological functions related to intercellular calcium (Ca2+) signaling. Quantitative analyzes of the spatial-temporal patterns of Ca2+ signal waves can provide valuable information for organ development and disease studies. Due to the nature of live imaging, the same wing pouch captured at different frames of an image sequence may incur considerable movement. As a result, registration, namely, accurately aligning the wing pouches across the whole image sequence, becomes a crucial step for Ca2+ spatial-temporal analysis. Since wing pouches in different image frames exhibit extensive intensity oscillations due to Ca2+ signaling dynamics, commonly used multimodal non-rigid registration methods fail to achieve satisfactory results. In this paper, we develop a new two-phase non-rigid registration approach to automatically register wing pouches in image sequences. First, we conduct segmentation of the region of interest (i.e., wing pouches) using a deep neural network, followed by a boundary refinement post processing. Second, we apply B-spline based registration to obtain an optimal transformation and align wing pouches across the whole image sequence. Evaluated using both synthetic data and real wing pouch data, our method considerably outperforms the state-of-the-art non-rigid registration algorithms and achieves good quality results for wing pouch image sequence analysis.