It sometimes happens (for instance in case control studies) that a classifier is trained on a data set that does not reflect the true a priori probabilities of the target classes on real-world data. This may have a negative effect on the classification accuracy obtained on the real-world data set, especially when the classifier's decisions are based on the a posteriori probabilities of class membership. Indeed, in this case, the trained classifier provides estimates of the a posteriori probabilities that are not valid for this real-world data set (they rely on the a priori probabilities of the training set). Applying the classifier as is (without correcting its outputs with respect to these new conditions) on this new data set may thus be suboptimal. In this note, we present a simple iterative procedure for adjusting the outputs of the trained classifier with respect to these new a priori probabilities without having to refit the model, even when these probabilities are not known in advance. As a by-product, estimates of the new a priori probabilities are also obtained. This iterative algorithm is a straightforward instance of the expectation-maximization (EM) algorithm and is shown to maximize the likelihood of the new data. Thereafter, we discuss a statistical test that can be applied to decide if the a priori class probabilities have changed from the training set to the real-world data. The procedure is illustrated on different classification problems involving a multilayer neural network, and comparisons with a standard procedure for a priori probability estimation are provided. Our original method, based on the EM algorithm, is shown to be superior to the standard one for a priori probability estimation. Experimental results also indicate that the classifier with adjusted outputs always performs better than the original one in terms of classification accuracy, when the a priori probability conditions differ from the training set to the real-world data. The gain in classification accuracy can be significant.

Background: Post-mortem studies can provide important information for understanding new diseases and small autopsy case series have already reported different findings in COVID-19 patients. Methods: We evaluated whether some specific post-mortem features are observed in these patients and if these changes are related to the presence of the virus in different organs. Complete macroscopic and microscopic autopsies were performed on different organs in 17 COVID-19 non-survivors. Presence of SARS-CoV-2 was evaluated with immunohistochemistry (IHC) in lung samples and with real-time reverse-transcription polymerase chain reaction (RT-PCR) test in the lung and other organs. Results: Pulmonary findings revealed early-stage diffuse alveolar damage (DAD) in 15 out of 17 patients and microthrombi in small lung arteries in 11 patients. Late-stage DAD, atypical pneumocytes, and/or acute pneumonia were also observed. Four lung infarcts, two acute myocardial infarctions, and one ischemic enteritis were observed. There was no evidence of myocarditis, hepatitis, or encephalitis. Kidney evaluation revealed the presence of hemosiderin in tubules or pigmented casts in most patients. Spongiosis and vascular congestion were the most frequently encountered brain lesions. No specific SARS-CoV-2 lesions were observed in any organ. IHC revealed positive cells with a heterogeneous distribution in the lungs of 11 of the 17 (65%) patients; RT-PCR yielded a wide distribution of SARS-CoV-2 in different tissues, with 8 patients showing viral presence in all tested organs (i.e., lung, heart, spleen, liver, colon, kidney, and brain). Conclusions: In conclusion, autopsies revealed a great heterogeneity of COVID-19-associated organ injury and the remarkable absence of any specific viral lesions, even when RT-PCR identified the presence of the virus in many organs.

In this paper, we propose a combination of mean-shift-based tracking processes to establish migrating cell trajectories through in vitro phase-contrast video microscopy. After a recapitulation on how the mean-shift algorithm permits efficient object tracking we describe the proposed extension and apply it to the in vitro cell tracking problem. In this application, the cells are unmarked (i.e., no fluorescent probe is used) and are observed under classical phase-contrast microscopy. By introducing an adaptive combination of several kernels, we address several problems such as variations in size and shape of the tracked objects (e.g., those occurring in the case of cell membrane extensions), the presence of incomplete (or noncontrasted) object boundaries, partially overlapping objects and object splitting (in the case of cell divisions or mitoses). Comparing the tracking results automatically obtained to those generated manually by a human expert, we tested the stability of the different algorithm parameters and their effects on the tracking results. We also show how the method is resistant to a decrease in image resolution and accidental defocusing (which may occur during long experiments, e.g., dozens of hours). Finally, we applied our methodology on cancer cell tracking and showed that cytochalasin-D significantly inhibits cell motility.

Cancer cell motility and invasion are critical targets for anticancer therapeutics. Whereas in vitro models could be designed for rapid screening with a view to investigate these targets, careful consideration must be given to the construction of appropriate model systems. Most investigations focus on two-dimensional (2-D) assays despite the fact that increasing evidence suggests that migration across rigid and planar substrates fails to recapitulate in vivo behavior. In contrast, few systems enable three-dimensional (3-D) cell migration to be quantitatively analyzed. We previously developed a digital holographic microscope (DHM) working in transmission with a partially spatial coherence source. This configuration avoids the noise artifacts of laser illumination and makes possible the direct recording of information on the 3-D structure of samples consisting of multiple objects embedded in scattering media, such as cell cultures in matrix gels. The software driving our DHM system is equipped with a time-lapse ability that enables the 3-D trajectories of living cells to be reconstituted and quantitatively analyzed.