In recent years, three-dimensional mesoscopic imaging has gained significant importance in life sciences for fundamental studies at the whole-organ level. In this manuscript, we present an optical projection tomography (OPT) method designed for imaging of the intact mouse brain. The system features an isotropic resolution of ~50 µm and an acquisition time of four to eight minutes, using a 3-day optimized clearing protocol. Imaging of the brain autofluorescence in 3D reveals details of the neuroanatomy, while the use of fluorescent labels displays the vascular network and amyloid deposition in 5xFAD mice, an important model of Alzheimer's disease (AD). Finally, the OPT images are compared with histological slices.

Optical projection tomography (OPT) is a powerful tool for three-dimensional imaging of mesoscopic biological samples with great use for biomedical phenotyping studies. We present a fluorescent OPT platform that enables direct visualization of biological specimens and processes at a centimeter scale with high spatial resolution, as well as fast data throughput and reconstruction. We demonstrate nearly isotropic sub-28 µm resolution over more than 60 mm3 after reconstruction of a single acquisition. Our setup is optimized for imaging the mouse gut at multiple wavelengths. Thanks to a new sample preparation protocol specifically developed for gut specimens, we can observe the spatial arrangement of the intestinal villi and the vasculature network of a 3-cm long healthy mouse gut. Besides the blood vessel network surrounding the gastrointestinal tract, we observe traces of vasculature at the villi ends close to the lumen. The combination of rapid acquisition and a large field of view with high spatial resolution in 3D mesoscopic imaging holds an invaluable potential for gastrointestinal pathology research.

No abstract

The growth of data throughput in optical microscopy has triggered the extensive use of supervised learning (SL) models on compressed datasets for automated analysis. Investigating the effects of image compression on SL predictions is therefore pivotal to assess their reliability, especially for clinical use. We quantify the statistical distortions induced by compression through the comparison of predictions on compressed data to the raw predictive uncertainty, numerically estimated from the raw noise statistics measured via sensor calibration. Predictions on cell segmentation parameters are altered by up to 15% and more than 10 standard deviations after 16-to-8 bits pixel depth reduction and 10:1 JPEG compression. JPEG formats with higher compression ratios show significantly larger distortions. Interestingly, a recent metrologically accurate algorithm, offering up to 10:1 compression ratio, provides a prediction spread equivalent to that stemming from raw noise. The method described here allows to set a lower bound to the predictive uncertainty of a SL task and can be generalized to determine the statistical distortions originated from a variety of processing pipelines in AI-assisted fields.

The growth of data throughput in optical microscopy has triggered the extensive use of supervised learning (SL) models on compressed datasets for automated analysis. Investigating the effects of image compression on SL predictions is therefore pivotal to assess their reliability, especially for clinical use.We quantify the statistical distortions induced by compression through the comparison of predictions on compressed data to the raw predictive uncertainty, numerically estimated from the raw noise statistics measured via sensor calibration. Predictions on cell segmentation parameters are altered by up to 15% and more than 10 standard deviations after 16-to-8 bits pixel depth reduction and 10:1 JPEG compression. JPEG formats with higher compression ratios show significantly larger distortions. Interestingly, a recent metrologically accurate algorithm, offering up to 10:1 compression ratio, provides a prediction spread equivalent to that stemming from raw noise. The method described here allows to set a lower bound to the predictive uncertainty of a SL task and can be generalized to determine the statistical distortions originated from a variety of processing pipelines in AI-assisted fields.