Improving disease classification performance and explainability of deep learning models in radiology with heatmap generators.

As deep learning is widely used in the radiology field, the explainability of Artificial Intelligence (AI) models is becoming increasingly essential to gain clinicians' trust when using the models for diagnosis. In this research, three experiment sets were conducted with a U-Net architecture to improve the disease classification performance while enhancing the heatmaps corresponding to the model's focus through incorporating heatmap generators during training. All experiments used the dataset that contained chest radiographs, associated labels from one of the three conditions ["normal", "congestive heart failure (CHF)", and "pneumonia"], and numerical information regarding a radiologist's eye-gaze coordinates on the images. The paper that introduced this dataset developed a U-Net model, which was treated as the baseline model for this research, to show how the eye-gaze data can be used in multi-modal training for explainability improvement and disease classification. To compare the classification performances among this research's three experiment sets and the baseline model, the 95% confidence intervals (CI) of the area under the receiver operating characteristic curve (AUC) were measured. The best method achieved an AUC of 0.913 with a 95% CI of [0.860, 0.966]. "Pneumonia" and "CHF" classes, which the baseline model struggled the most to classify, had the greatest improvements, resulting in AUCs of 0.859 with a 95% CI of [0.732, 0.957] and 0.962 with a 95% CI of [0.933, 0.989], respectively. The decoder of the U-Net for the best-performing proposed method generated heatmaps that highlight the determining image parts in model classifications. These predicted heatmaps, which can be used for the explainability of the model, also improved to align well with the radiologist's eye-gaze data. Hence, this work showed that incorporating heatmap generators and eye-gaze information into training can simultaneously improve disease classification and provide explainable visuals that align well with how the radiologist viewed the chest radiographs when making diagnosis.

Streptococcus panodentis sp. nov. from the oral cavities of chimpanzees.

Three strains TKU9, TKU49 and TKU50(T) , were isolated from the oral cavities of chimpanzees (Pan troglodytes). The isolates were all gram-positive, facultative anaerobic cocci that lacked catalase activity. Analysis of partial 16S rRNA gene sequences showed that the most closely related species was Streptococcus infantis (96.7%). The next most closely related species to the isolates were S. rubneri, S. mitis, S. peroris and S. australis (96.6 to 96.4%). Based on the rpoB and gyrB gene sequences, TKU50(T) was clustered with other member of the mitis group. Enzyme activity and sugar fermentation patterns differentiated this novel bacterium from other members of the mitis group streptococci. The DNA G + C content of strain TKU50(T) was 46.7 mol%, which is the highest reported value for members of the mitis group (40-46 mol%). On the basis of the phenotypic characterization, partial 16S rRNA gene and sequences data for two housekeeping gene (gyrB and rpoB), we propose a novel taxa, S. panodentis for TKU 50(T) (type strain = CM 30579(T) = DSM 29921(T) ), for these newly described isolates.

Emergence of infectious malignant thrombocytopenia in Japanese macaques (Macaca fuscata) by SRV-4 after transmission to a novel host.

We discovered a lethal hemorrhagic syndrome arising from severe thrombocytopenia in Japanese macaques kept at the Primate Research Institute, Kyoto University. Extensive investigation identified that simian retrovirus type 4 (SRV-4) was the causative agent of the disease. SRV-4 had previously been isolated only from cynomolgus macaques in which it is usually asymptomatic. We consider that the SRV-4 crossed the so-called species barrier between cynomolgus and Japanese macaques, leading to extremely severe acute symptoms in the latter. Infectious agents that cross the species barrier occasionally amplify in virulence, which is not observed in the original hosts. In such cases, the new hosts are usually distantly related to the original hosts. However, Japanese macaques are closely related to cynomolgus macaques, and can even hybridize when given the opportunity. This lethal outbreak of a novel pathogen in Japanese macaques highlights the need to modify our expectations about virulence with regards crossing species barriers.

Intracranial arachnoid cysts in a chimpanzee (Pan troglodytes).

An intracranial arachnoid cyst was detected in a 32-year-old, 44.6-kg, female chimpanzee at the Primate Research Institute, Kyoto University. Magnetic resonance imaging (MRI) and computed tomography (CT) were performed and the cognitive studies in which she participated were reviewed. MRI revealed that the cyst was present in the chimpanzee's right occipital convexity, and was located in close proximity to the posterior horn of the right lateral ventricle without ventriculomegaly. CT confirmed the presence of the cyst and no apparent signs indicating previous skull fractures were found. The thickness of the mandible was asymmetrical, whereas the temporomandibular joints and dentition were symmetrical. She showed no abnormalities in various cognitive studies since she was 3 years old, except a different behavioural pattern during a recent study, indicating a possible visual field defect. Detailed cognitive studies, long-term observation of her physical condition and follow-up MRI will be continued.

Characterization of simian T-cell leukemia virus type 1 in naturally infected Japanese macaques as a model of HTLV-1 infection.

BACKGROUND: Human T-cell leukemia virus type 1 (HTLV-1) causes chronic infection leading to development of adult T-cell leukemia (ATL) and inflammatory diseases. Non-human primates infected with simian T-cell leukemia virus type 1 (STLV-1) are considered to constitute a suitable animal model for HTLV-1 research. However, the function of the regulatory and accessory genes of STLV-1 has not been analyzed in detail. In this study, STLV-1 in naturally infected Japanese macaques was analyzed. RESULTS: We identified spliced transcripts of STLV-1 corresponding to HTLV-1 tax and HTLV-1 bZIP factor (HBZ). STLV-1 Tax activated the NFAT, AP-1 and NF-κB signaling pathways, whereas STLV-1 bZIP factor (SBZ) suppressed them. Conversely, SBZ enhanced TGF-ß signaling and induced Foxp3 expression. Furthermore, STLV-1 Tax activated the canonical Wnt pathway while SBZ suppressed it. STLV-1 Tax enhanced the viral promoter activity while SBZ suppressed its activation. Then we addressed the clonal proliferation of STLV-1⁺ cells by massively sequencing the provirus integration sites. Some clones proliferated distinctively in monkeys with higher STLV-1 proviral loads. Notably, one of the monkeys surveyed in this study developed T-cell lymphoma in the brain; STLV-1 provirus was integrated in the lymphoma cell genome. When anti-CCR4 antibody, mogamulizumab, was administered into STLV-1-infected monkeys, the proviral load decreased dramatically within 2 weeks. We observed that some abundant clones recovered after discontinuation of mogamulizumab administration. CONCLUSIONS: STLV-1 Tax and SBZ have functions similar to those of their counterparts in HTLV-1. This study demonstrates that Japanese macaques naturally infected with STLV-1 resemble HTLV-1 carriers and are a suitable model for the investigation of persistent HTLV-1 infection and asymptomatic HTLV-1 carrier state. Using these animals, we verified that mogamulizumab, which is currently used as a drug for relapsed ATL, is also effective in reducing the proviral load in asymptomatic individuals.

Efficient in vivo depletion of CD8(+) T lymphocytes in common marmosets by novel CD8 monoclonal antibody administration.

In order to directly demonstrate the roles of CD8(+) T lymphocytes in non-human primates, in vivo depletion of the CD8(+) T cells by administration of a CD8-specific monoclonal antibody (mAb) is one of the crucial techniques. Recently, the common marmoset (Callithrix jacchus), which is classified as a New World monkey, has been shown useful as an experimental animal model for various human diseases such as multiple sclerosis, Parkinson's disease and a number of infectious diseases. Here we show that an anti-marmoset CD8 mAb 6F10, which we have recently established, efficiently depletes the marmoset CD8(+) T lymphocytes in vivo, i.e., the administration of 6F10 induces drastic and specific reduction in the ratio of the CD8(+) T cell subset for at least three weeks or longer. Our finding will help understand the pivotal role of CD8(+) T cells in vivo in the control of human diseases.

Streptococcus troglodytae sp. nov., from the chimpanzee oral cavity.

Six strains, TKU 25, TKU 28, TKU 30, TKU 31(T), TKU 33 and TKU 34, were isolated from the oral cavity of a chimpanzee (Pan troglodytes). Colonies of strains grown on Mitis-Salivarius agar were similar in morphology to that of Streptococcus mutans. The novel strains were Gram-stain-positive, facultatively anaerobic cocci that lacked catalase activity. Analysis of the partial 16S rRNA gene sequences of these isolates showed that the most closely related strain was the type strain of S. mutans (96.4 %). The next closely related strains to the isolates were the type strains of Streptococcus devriesei (94.5 %) and Streptococcus downei (93.9 %). These isolates could be distinguished from S. mutans by inulin fermentation and alkaline phosphatase activity (API ZYM system). The peptidoglycan type of the novel isolates was Glu-Lys-Ala(3). Strains were not susceptible to bacitracin. On the basis of phenotypic characterization, partial 16S rRNA gene and two housekeeping gene (groEL and sodA) sequence data, we propose a novel taxon, Streptococcus troglodytae sp. nov.; the type strain is TKU 31(T) ( = JCM 18038(T) = DSM 25324(T)).