**4. Discussion**

The presented results demonstrate that motion rejection in OAT can effectively be accomplished prior to image reconstruction. This represents a significant advantage with respect to previously reported motion rejection approaches based on auto-correlation of a sequence of reconstructed images [31], which are afflicted with excessive memory and post-processing requirements. The suggested method was successfully validated with data acquired by two- and three-dimensional imaging systems. However, motion rejection was more effective in the case of volumetric SVOT scans. In particular, it benefited from both amplitude increase of 10% to 30% and improvement in the visibility of fine details, whereas images from the cross-sectional imaging system yielded a lower amplitude increase (~10%) and minor improvement in the visibility of structures. The reason behind the reduced performance of motion rejection in cross-sectional imaging may be ascribed to the fact that breathing-associated movements are not limited to a single plane, while in-plane motion is mainly detected in the signals. Yet, although the differences between selected- and all-frames cross-sectional images were minor, it was possible to quantify them by utilizing a QI based distortion measures. Notably, such distortion artifacts affect almost exclusively the edges of large structures. In spite of the fact that standard frame averaging in cross-sectional imaging may yield qualitatively comparable results, reliable rejection of 20% to 31% motion-affected frames by the algorithm may turn crucial for quantitative analyses of high resolution data, e.g., involving spectral unmixing of fine structures.

It is also important to take into account that breathing characteristics may differ from one animal to another due to age, health, size, sex or strain. All these factors affect the resilience of the animal to the experimental setup, the feasible depth of anesthesia and the overall duration of the experiment [38]. It was previously reported that mice under 2% isoflurane anesthesia have an average respiratory rate of 44 ± 9 breaths/min [39], where the breathing rhythm is characterized by pauses between breaths longer than the breaths themselves. As a result, the majority of the frames are static, i.e., not affected by motion. Herein, we relied on such prior knowledge of the characteristic respiratory rate and breathing rhythm to establish a rejection criterion for the clustered motion (rejected) frames. Likewise, other criteria independent of these factors may alternatively be implemented.

In conclusion, the developed motion rejection methodology can benefit numerous optoacoustic imaging methods relying on multi-frame image analysis, such as scanning-based tomography or spectroscopic imaging systems like the MSOT. It may also find applicability in handheld clinical imaging [40,41], where motion can hinder accurate signal quantification and interpretation of longitudinal and spectroscopic data.

**Author Contributions:** Conceptualization, A.R., X.L.D.-B. and D.R.; methodology, A.R.; software, A.R.; validation, A.R. and N.D.; formal analysis, A.R.; investigation, A.R.; resources, D.R.; data curation, A.R. and X.L.D.-B.; writing—original draft preparation, A.R., X.L.D.-B. and D.R.; writing—review and editing, A.R., X.L.D.-B. and D.R.; visualization, A.R.; supervision, D.R.; project administration, D.R.; funding acquisition, D.R.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors wish to thank M. Reiss for his support with the measurements and handling of animals.

**Conflicts of Interest:** The authors declare no conflict of interest.
