Search

Deep learning models have become state-of-the-art in many areas, ranging from computer vision to agriculture research. However, concerns have been raised with respect to the transparency of their decisions, especially in the image domain. In this regard, Explainable Artificial Intelligence has been gaining popularity in recent years. The ProtoPNet model, which breaks down an image into prototypes and uses evidence gathered from the prototypes to classify an image, represents an appealing approach. Still, questions regarding its effectiveness arise when the application domain changes from real-world natural images to gray-scale medical images. This work explores the applicability of prototypical part learning in medical imaging by experimenting with ProtoPNet on a breast masses classification task. The two considered aspects were the classification capabilities and the validity of explanations. We looked for the optimal model's hyperparameter configuration via a random search. We trained the model in a five-fold CV supervised framework, with mammogram images cropped around the lesions and ground-truth labels of benign/malignant masses. Then, we compared the performance metrics of ProtoPNet to that of the corresponding base architecture, which was ResNet18, trained under the same framework. In addition, an experienced radiologist provided a clinical viewpoint on the quality of the learned prototypes, the patch activations, and the global explanations. We achieved a Recall of 0.769 and an area under the receiver operating characteristic curve of 0.719 in our experiments. Even though our findings are non-optimal for entering the clinical practice yet, the radiologist found ProtoPNet's explanations very intuitive, reporting a high level of satisfaction. Therefore, we believe that prototypical part learning offers a reasonable and promising trade-off between classification performance and the quality of the related explanation.

The Signal & Images Laboratory is an interdisciplinary research group in computer vision, signal analysis, intelligent vision systems and multimedia data understanding. It is part of the Institute of Information Science and Technologies (ISTI) of the National Research Council of Italy (CNR). This report accounts for the research activities of the Signal and Images Laboratory of the Institute of Information Science and Technologies during the year 2021.

The discovery of several sources of hepatic progenitors in extra-hepatic organs and tissues, both in animal models and in humans, supports opportunities to isolate, grow and expand them in vitro. Microenvironment factors involved in regulating proliferation and commitment of liver cell precursors have been identified and better characterization of liver stem cell pathobiology would greatly improve the understanding of liver differentiation/ regeneration processes, especially those leading to hepatocarcinogenesis. The goal of these researches has been to discover the most available, suitable and easy-to-use pluripotent stem cells (PSC) sources for cell-based therapies in genetic and acquired liver diseases, therapies which to date have required liver transplantation. This report reviews the efforts, so far, to either investigate the presence of PSC in hepatic and extra-hepatic districts or evaluate their capacity to differentiate in vitro and to restore in vivo liver function.

Iacovacci et al. [(1997a) Research in Virology 148:147-151] described that the euploid diploid cells, of the normal human bone marrow-derived lymphoblastoid B-cell line TO.FE., are susceptible to hepatitis C virus (HCV) infection and support long term virus production. Transmission electron microscopy described some steps of HCV replication cycle in this in vitro infected cellular system [Serafino et al. (1997) Research in Virology 148:153-159]. In the present study, in order to identify the intracellular sites involved in HCV replication, the ultrastructural changes associated with infection in TO.FE. cells were correlated with the subcellular localisation of structural and nonstructural viral proteins. Transmission electron microscopy and confocal microscopy data indicate that these viral proteins appeared located in the Golgi apparatus and endoplasmic reticulum, suggesting an active involvement of these compartments in viral assembly and morphogenesis. Furthermore, transmission and scanning electron microscopic observations on cultures infected chronically support the hypothesis that these cellular compartments may serve as starting sites of the morphological changes associated to viral infection and replication, leading to cell-cell fusion, syncytia formation, and finally lysis of infected cells and virus release.

It has been suggested that an atypical course of primary infection by EBV and the reactivation of EBV infection in transplanted patients may induce hepatitis. We explored the possibility to dissect the infectious activity from the ability to promote B lymphocyte proliferation in vivo by injecting in nu/nu mice a low number (2 x 10(6)-0.05 x 10(6)) of cells from CE a normal human bone marrow-derived B cell line. This line carries an endogenous EBV in episomal and linear forms. Twenty nu/nu mice were inoculated subcutaneously with the B cell line CE and a matched group with the cell line RAG obtained by EBV in vitro infection of normal human peripheral blood. The mice injected with the CE line did not develop a lymphoproliferative disease, but 5 of them displayed typical histopathological lesions of chronic hepatitis without involvement of other organs. Similar results were obtained in 2 out of 20 animals in the RAG group. A close association between liver lesions and a previous EBV infection, by putative circulating B lymphoblastoid cells releasing their EBV, was established by PCR and by in situ hybridization with BamHI "W" DNA probe. This latter probe detected the presence of about 15% of positive cells only in affected livers. In addition, the rare detection in some hepatocytes of "A" type Cowdry bodies would suggest the occurrence of continuous EBV replication although at a very low level. These data show that we succeeded in dissecting the infectious from the proliferative activity of the endogenous EBV carrier CE cell line. This provides in addition a promising model for chronic EBV-associated hepatitis.

Six putative novel human papillomavirus (HPV) types were detected by using general primers for a conserved L1 HPV region in patients examined in gynecologic centers. One of the isolates, detected in samples from 4 patients with koilocytic atypia at cervical cytology (3 of whom were also infected with human immunodeficiency virus type 1), was completely sequenced, identified as a new HPV genotype, and designated candidate HPV87 (candHPV87) by the Reference Center for Human Papillomavirus. candHPV87 shows the classic HPV genome organization and the absence of a functional E5 coding region. Phylogenetic analysis documented that the candHPV87 genome clusters within the A3 group of HPVs, together with HPV61, HPV72, HPV83, HPV84 and candHPV86, which have been completely sequenced, and a number of other putative novel genotypes (two of which are described in this work), which have been partially characterized. To address the growth-enhancing potential of candHPV87, the E6 and E7 putative coding regions were cloned and expressed in tissue cultures. The data indicate that both proteins stimulate cell division in tissue cultures more than those of low-risk HPVs, though not as much as those of HPV16. Taken together, the clinical, molecular, and biological data suggest that the novel papillomavirus characterized in the present study is a low- to intermediate-risk HPV.