Tuning Vision-Language Models with Multiple Prototypes Clustering.

Benefiting from advances in large-scale pre-training, foundation models, have demonstrated remarkable capability in the fields of natural language processing, computer vision, among others. However, to achieve expert-level performance in specific applications, such models often need to be fine-tuned with domain-specific knowledge. In this paper, we focus on enabling vision-language models to unleash more potential for visual understanding tasks under few-shot tuning. Specifically, we propose a novel adapter, dubbed as lusterAdapter, which is based on trainable multiple prototypes clustering algorithm, for tuning the CLIP model. It can not only alleviate the concern of catastrophic forgetting of foundation models by introducing anchors to inherit common knowledge, but also improve the utilization efficiency of few annotated samples via bringing in clustering and domain priors, thereby improving the performance of few-shot tuning. We have conducted extensive experiments on 11 common classification benchmarks. The results show our method significantly surpasses the original CLIP and achieves state-of-the-art (SOTA) performance under all benchmarks and settings. For example, under the 16-shot setting, our method exhibits a remarkable improvement over the original CLIP by 19.6%, and also surpasses TIP-Adapter and GraphAdapter by 2.7% and 2.2%, respectively, in terms of average accuracy across the 11 benchmarks. Code is available at https://github.com/uyzhang/Cluster-Adapter.

SN 2 eRF: A Framework for Neural Radiance Fields given Sparse and Noisy Poses.

Neural Radiance Fields (NeRFs) have shown impressive capabilities in synthesizing photorealistic novel views. However, their application to room-size scenes is limited by the requirement of several hundred views with accurate poses for training. To address this challenge, we propose SN 2 eRF, a framework which can reconstruct the neural radiance field with significantly fewer views and noisy poses by exploiting multiple priors. Our key insight is to leverage both multi-view and monocular priors to constrain the optimization of NeRF in the setting of sparse and noisy pose inputs. Specifically, we extract and match key points to constrain pose optimization and use Ray Transformer with a monocular depth estimator to provide dense depth prior for geometry optimization. Benefiting from these priors, our approach achieves state-of-the-art accuracy in novel view synthesis for indoor room scenarios.

Optical genome mapping improves the accuracy of classification, risk stratification, and personalized treatment strategies for patients with acute myeloid leukemia.

Cytogenomic characterization is crucial for the classification and risk stratification of acute myeloid leukemia (AML), thereby facilitating therapeutic decision-making. We examined the clinical utility of optical genome mapping (OGM) in 159 AML patients (103 newly diagnosed and 56 refractory/relapsed), all of whom also underwent chromosomal banding analysis (CBA), fluorescence in situ hybridization, and targeted next-generation sequencing. OGM detected nearly all clinically relevant cytogenetic abnormalities that SCG identified with >99% sensitivity, provided the clonal burden was above 20%. OGM identified additional cytogenomic aberrations and/or provided information on fusion genes in 77 (48%) patients, including eight patients with normal karyotypes and four with failed karyotyping. The most common additional alterations identified by OGM included chromoanagenesis (n = 23), KMT2A partial tandem duplication (n = 11), rearrangements involving MECOM (n = 7), NUP98 (n = 2), KMT2A (n = 2), JAK2 (n = 2), and other gene fusions in 17 patients, with 10 showing novel fusion gene partners. OGM also pinpointed fusion genes in 17 (11%) patients where chromosomal rearrangements were concurrently detected by OGM and CBA. Overall, 24 (15%) aberrations were identified exclusively by OGM and had the potential to alter AML classification, risk stratification, and/or clinical trial eligibility. OGM emerges as a powerful tool for identifying fusion genes and detecting subtle or cryptic cytogenomic aberrations that may otherwise remain undetectable by CBA.

Learning Virtual View Selection for 3D Scene Semantic Segmentation.

2D-3D joint learning is essential and effective for fundamental 3D vision tasks, such as 3D semantic segmentation, due to the complementary information these two visual modalities contain. Most current 3D scene semantic segmentation methods process 2D images "as they are", i.e., only real captured 2D images are used. However, such captured 2D images may be redundant, with abundant occlusion and/or limited field of view (FoV), leading to poor performance for the current methods involving 2D inputs. In this paper, we propose a general learning framework for joint 2D-3D scene understanding by selecting informative virtual 2D views of the underlying 3D scene. We then feed both the 3D geometry and the generated virtual 2D views into any joint 2D-3D-input or pure 3D-input based deep neural models for improving 3D scene understanding. Specifically, we generate virtual 2D views based on an information score map learned from the current 3D scene semantic segmentation results. To achieve this, we formalize the learning of the information score map as a deep reinforcement learning process, which rewards good predictions using a deep neural network. To obtain a compact set of virtual 2D views that jointly cover informative surfaces of the 3D scene as much as possible, we further propose an efficient greedy virtual view coverage strategy in the normal-sensitive 6D space, including 3-dimensional point coordinates and 3-dimensional normal. We have validated our proposed framework for various joint 2D-3D-input or pure 3D-input based deep neural models on two real-world 3D scene datasets, i.e., ScanNet v2 and S3DIS, and the results demonstrate that our method obtains a consistent gain over baseline models and achieves new top accuracy for joint 2D and 3D scene semantic segmentation. Code is available at https://github.com/smy-THU/VirtualViewSelection.

Morphology, immunophenotype, and suggested diagnostic criteria of TCL1 family-negative T-prolymphocytic leukemia.

OBJECTIVES: We sought to investigate the morphologic and immunophenotypic characteristics of TCL1 family-negative T-cell prolymphocytic leukemia (T-PLL). METHODS: Twenty cases of TCL1 family-negative T-PLL were studied. RESULTS: The doubling time of leukemic cells ranged from less than 2 days to more than 5 years, with a median of 5.5 months. Leukemic cells were small to medium-sized, with round to irregular nuclei, variably condensed chromatin, and small amounts of agranular cytoplasm. A visible nucleolus was identified in 11 (55%) cases. Cytoplasmic blebs/protrusions were identified in all cases, but their occurrence was highly variable from case to case. Bone marrow biopsy showed an interstitial pattern in 90% of cases and a diffuse pattern in the remaining 10% of cases. Flow cytometric immunophenotypic analysis showed that the leukemic cells in all cases were CD4 positive; 3 (15%) also showed concurrent CD8 expression. All cases were positive for CD2 and CD5. Surface CD3 and CD7 were positive in 19 of 20 (95%) cases, and all CD3-positive cases expressed the T-cell receptor αß. Compared with prototypic T-PLL cases, these 2 groups shared many immunophenotypic findings, except CD8 and CD26, both of which were more commonly expressed in prototypic T-PLL cases. CONCLUSIONS: TCL1 family-negative T-PLL cases have morphologic and immunophenotypic features that are similar to prototypic T-PLL. They are characterized by neoplastic proliferation of small to medium-sized mature T cells with CD4-positive T-cell receptor αß phenotype. Tumor cells frequently maintain pan-T antigen expression. Recognizing these morphologic and immunophenotypic features will aid in accurately diagnosing this rare subset of T-PLL.

Outcome of 3q26.2/MECOM rearrangements in chronic myeloid leukemia.

STUDY AIMS: To evaluate the outcomes of patients with 3q26.2/MECOM-rearranged chronic myeloid leukemia (CML). METHODS: We reviewed consecutive adult patients with 3q26.2/MECOM-rearranged CML between January 1, 1998 and February 16, 2023. Rearrangements of 3q26.2/MECOM were confirmed by conventional cytogenetics, and fluorescence in situ hybridization starting in 2015. RESULTS: We identified 55 patients with MECOM-rearranged CML, including 23 in chronic phase (CP) or accelerated phase (AP) and 32 in blast phase (BP). Nine patients (16%) achieved a major cytogenetic response (MCyR) or deeper. At a median follow-up of 89 months, median survival was 14 months. The 5-year survival rate was 19% overall, 23% in CML-CP/AP, and 15% in CML-BP. In the 6-month landmark analysis, the 5-year survival rate was 41% for allogeneic stem cell transplantation (allo-SCT) recipients versus 17% for non-recipients (P = 0.050). Multivariate analysis showed that the percentage of marrow blasts and achievement of MCyR or deeper could predict survival. CONCLUSION: Outcomes of 3q26.2/MECOM-rearranged CML are poor despite the availability of multiple BCR::ABL1 tyrosine kinase inhibitors (TKIs). Third-generation TKIs in combination with novel agents and possible allo-SCT could be considered given the poor outcomes and resistance to second-generation TKIs.

Integrative immunophenotypic and genetic characterization of acute myeloid leukemia with CBFB rearrangement.

OBJECTIVES: We sought to characterize the immunophenotype of acute myeloid leukemia (AML) with CBFB rearrangement and correlate the results with cytogenetic and molecular data. METHODS: Sixty-one cases of AML with CBFB rearrangement were evaluated. RESULTS: The sample population consisted of 33 men and 28 women, with a median age of 49 years. Flow cytometry immunophenotypic analysis showed that myeloblasts were positive for CD34 and CD117 in all cases, and myeloperoxidase was positive in 52 of 55 (95%) cases. The most common abnormalities included decreased CD38 in 90%, increased CD13 in 85%, increased CD123 in 84%, and decreased HLA-DR in 84% of cases. Monocytes were increased, with a mature immunophenotype, and accounted for 23.7% of total cells. Among 60 cases with available karyotype, inv(16)(p13.1q22) was most common in 50 (83%) cases, followed by t(16;16) (p13.1;q22) in 6 (10%). Type A CBFB::MYH11 transcript was most common, detected in 84% of cases. Mutational analysis showed mutations of NRAS in 37%, FLT3 in 25%, and KIT in 24% of cases. Comparing cases with type A vs non-type A transcripts, blasts in type A cases more frequently exhibited CD64 positivity and increased CD13 levels while showing a lower frequency of CD7 and CD56 expression. Trisomy 22 and mutations in KIT, NF1, and TET2 were identified only in cases with type A transcript. CONCLUSIONS: Myeloblasts of AML with CBFB rearrangement are positive for CD34, CD117, and myeloperoxidase. These neoplasms most frequently carry inv(16)(p13.1q22) and type A fusion transcript. NRAS mutation was the most common mutation. Some immunophenotypic and genetic correlations occurred with different types of transcripts.

Optical genomic mapping is a helpful tool for detecting CCND1 rearrangements in CD5-negative small B-cell lymphoma: Two cases of leukemic non-nodal mantle cell lymphoma.

Optical genome mapping (OGM) is a new DNA-based technology which provides comprehensive examination of the entire genome. We report two patients who presented with splenomegaly and leukocytosis with lymphocytosis including villous lymphocytes. Neither patient had lymphadenopathy. Bone marrow evaluation showed involvement by small B-cell lymphoma in a sinusoidal and interstitial distribution, and immunophenotypic analysis showed that the neoplastic cells were positive for B-cell markers and cyclin D1 but were negative for SOX11 and CD5. Initially, the clinicopathologic features in both patients were thought to be suspicious for hairy cell leukemia variant or splenic marginal zone lymphoma. However, OGM detected CCND1 rearrangement: t(2;11)/IGK::CCND1 in one case and t(11;14)/IGH::CCND1 in the other case. These cases illustrate the valuable role OGM can play in establishing the diagnosis of MCL. Case 1 also contributes to the paucity of literature on the rare occurrence of IGK::CCND1 in MCL.

Cohort profile: Beijing Healthy Aging Cohort Study (BHACS).

The Beijing Healthy Aging Cohort Study (BHACS) was established to supplement the limited data of a large representative cohort of older people based on the general population and was designed to evaluate the prevalence, incidence, and natural history of cognitive decline, functional disability, and conventional vascular risk factors. The aim was to determine the evolution of these conditions by estimating the rates and determinants of progression and regression to adverse outcomes, including dementia, cardiovascular events, cancer, and all-cause death. It can therefore provide evidence to help policy makers develop better policies to promote healthy aging in China. BHACS consisted of three cohorts (BLSA, CCHS-Beijing, and BECHCS) in Beijing with a total population of 11 235 (6281 in urban and 4954 in rural areas) and an age range of 55 years or older (55-101 years) with a mean age of 70.35 ± 7.71 years (70.69 ± 7.62 years in urban and 69.92 ± 7.80 years in rural areas). BHACS-BLSA conducted the baseline survey in 2009 with a multistage stratification-random clustering procedure for people aged 55 years or older; BHACS-CCHS-Beijing conducted the baseline survey in 2013-2015 with a stratified multistage cluster random sampling method for people aged 55 years or older; and BHACS-BECHCS conducted the baseline survey in 2010-2014 with two-stage cluster random sampling method for people aged 60 years or older. Data were collected through questionnaires, physical measurements, and laboratory analyses. Topics covered by BHACS include a wide range of physical and mental health indicators, lifestyles and personal, family, and socio-economic determinants of health. There are no immediate plans to make the cohort data freely available to the public, but specific proposals for further collaboration are welcome. For further information and collaboration, please contact the corresponding author Yao He (e-mail: yhe301@x263.net).

Epstein-Barr virus-based prognostic model in nodular sclerosis classic Hodgkin lymphoma.

The role of Epstein-Barr virus (EBV) in lymphoma cells of nodular sclerosis classic Hodgkin lymphoma (NScHL) is controversial. Our aim was to explore this and establish a clinically feasible model for risk stratification. We interrogated data from 542 consecutive subjects with NScHL receiving ABVD therapy and demonstrated EBV-infection in their lymphoma cells with EBV-encoded small RNAs (EBERs) in situ hybridization. Subjects were divided into training and validation datasets. As data from the training dataset suggested EBERs-positivity was the only independent prognostic factor for both progression-free survival (PFS) and overall survival (OS), we developed corresponding prognostic models based on it. Our models showed excellent performance in both training and validation cohort. These data indicate the close association of EBV infection and the outcomes of persons with NScHL receiving ABVD. Additionally, our newly developed models should help physicians estimate prognosis and select individualized therapy.

Aleukemic Chronic Myeloid Leukemia Without Neutrophilia and Thrombocytosis: A Report From the BCR::ABL1 Pathology Group.

Chronic myeloid leukemia (CML) is characterized by leukocytosis with left-shifted neutrophilia, basophilia, eosinophilia, and variable thrombocytosis. However, extremely rare cases of patients with CML without significant leukocytosis and thrombocytosis (aleukemic phase [ALP] CML, or CML-ALP) have been reported. Due to its rarity and limited awareness, there remains a significant knowledge gap concerning the pathologic diagnosis, disease progression, and optimal patient management and outcomes. In this multi-institutional study, we investigated 31 patients with CML-ALP. Over half (54.8%) of patients had a history of or concurrent hematopoietic or nonhematopoietic malignancies. At time of diagnosis of CML-ALP, approximately 26.7% of patients exhibited neutrophilia, 56.7% had basophilia, and 13.3% showed eosinophilia. The median number of metaphases positive for t(9;22)(q34;q11.2) was 15, with a median of 38.5% of interphase nuclei positive for BCR::ABL1 by fluorescence in situ hybridization. The median BCR::ABL1 level was 26.14%. Remarkably, 14 (45.2%) patients were initially misdiagnosed or not diagnosed before karyotype or fluorescence in situ hybridization information for BCR::ABL1 became available. Twenty-five patients received tyrosine kinase inhibitors (TKIs). One patient developed blast crisis while on TKI treatment 8 months after initial diagnosis. With a median follow-up time of 46.1 months, 20 of 22 patients who received TKI therapy and had detailed follow-up information achieved complete cytogenetic remission or deeper, 15 achieved major molecular remission or deeper, and 10 achieved molecularly undetectable leukemia. In conclusion, given the frequent occurrence of prior or concurrent malignancies, aleukemic presentation, and low level of t(9;22)(q34;q11.2)/BCR::ABL1, misdiagnosis or delayed diagnosis is common among these patients. While these patients generally respond well to TKIs, rare patients may develop blastic transformation. It is therefore important for pathologists and hematologists to be aware of this highly unusual presentation of CML to ensure timely diagnosis and appropriate management.

Cytogenetic Profile in Monoclonal Gammopathy of Undetermined Significance, Smoldering and Symptomatic Multiple Myeloma: A Study of 1087 Patients with Highly Purified Plasma Cells.

The aim of this study was to examine the cytogenetic profiles of plasma cell neoplasms (PCNs) at various disease stages, encompassing 1087 patients with monoclonal gammopathy of undetermined significance (MGUS), smoldering multiple myeloma (SMM), newly diagnosed multiple myeloma (NDMM), and refractory/relapsed multiple myeloma (RRMM). Fluorescence in situ hybridization (FISH) analyses were conducted on highly purified plasma cell samples, revealing that 96% of patients exhibited at least one cytogenetic abnormality. The genomic complexity escalated from MGUS to SMM and further to NDMM and RRMM, largely driven by 1q gain, del(17p), MYC-rearrangement (MYC-R), del(1p), and tetraploidy. Elevated frequencies of high-risk cytogenetics (59%), 1q gain (44%), and del(17p) (23%), as well as the presence of subclones (48%), were particularly notable in RRMM cases. IGH::CCND1 was observed in 26% of the cases, with no apparent variations across races, ages, or disease groups. Concurrent chromosomal analysis with FISH revealed that the incidence of abnormal karyotypes was strongly correlated with the extent of neoplastic plasma cell infiltration, genomic complexity, and the presence of specific abnormalities like del(17p) and MYC-R. Approximately 98% of the cases with abnormal karyotypes were complex, with most featuring five or more abnormalities. Chromosome 1 structural abnormalities were the most prevalent, found in 65% of cases. The frequent presence of subclones and composite karyotypes underscored the genomic heterogeneity and instability in this cohort.

Cerebrospinal fluid oligoclonal bands in Chinese patients with multiple sclerosis: the prevalence and its association with clinical features.

Background: Cerebrospinal fluid oligoclonal band (CSF-OCB) is an established biomarker in diagnosing multiple sclerosis (MS), however, there are no nationwide data on CSF-OCB prevalence and its diagnostic performance in Chinese MS patients, especially in the virtue of common standard operation procedure (SOP). Methods: With a consensus SOP and the same isoelectric focusing system, we conducted a nationwide multi-center study on OCB status in consecutively, and recruited 483 MS patients and 880 non-MS patients, including neuro-inflammatory diseases (NID, n = 595) and non-inflammatory neurological diseases (NIND, n=285). Using a standardized case report form (CRF) to collect the clinical, radiological, immunological, and CSF data, we explored the association of CSF-OCB positivity with patient characters and the diagnostic performance of CSF-OCB in Chinese MS patients. Prospective source data collection, and retrospective data acquisition and statistical data analysis were used. Findings: 369 (76.4%) MS patients were OCB-positive, while 109 NID patients (18.3%) and 6 NIND patients (2.1%) were OCB-positive, respectively. Time from symptom onset to diagnosis was significantly shorter in OCB-positive than that in OCB-negative MS patients (13.2 vs 23.7 months, P=0.020). The prevalence of CSF-OCB in Chinese MS patients was significantly higher in high-latitude regions (41°-50°N)(P=0.016), and at high altitudes (>1000m)(P=0.025). The diagnostic performance of CSF-OCB differentiating MS from non-MS patients yielded a sensitivity of 76%, a specificity of 87%. Interpretation: The nationwide prevalence of CSF-OCB was 76.4% in Chinese MS patients, and demonstrated a good diagnostic performance in differentiating MS from other CNS diseases. The CSF-OCB prevalence showed a correlation with high latitude and altitude in Chinese MS patients.

Causal associations between sleep traits and brain structure: a bidirectional Mendelian randomization study.

BACKGROUND: Emerging evidence suggests bidirectional causal relationships between sleep disturbance and psychiatric disorders, but the underlying mechanisms remain unclear. Understanding the bidirectional causality between sleep traits and brain imaging-derived phenotypes (IDPs) will help elucidate the mechanisms. Although previous studies have identified a range of structural differences in the brains of individuals with sleep disorders, it is still uncertain whether grey matter (GM) volume alterations precede or rather follow from the development of sleep disorders. RESULTS: After Bonferroni correction, the forward MR analysis showed that insomnia complaint remained positively associated with the surface area (SA) of medial orbitofrontal cortex (ß, 0.26; 95% CI, 0.15-0.37; P = 5.27 × 10-6). In the inverse MR analysis, higher global cortical SA predisposed individuals less prone to suffering insomnia complaint (OR, 0.89; 95%CI, 0.85-0.94; P = 1.51 × 10-5) and short sleep (≤ 6 h; OR, 0.98; 95%CI, 0.97-0.99; P = 1.51 × 10-5), while higher SA in posterior cingulate cortex resulted in a vulnerability to shorter sleep durations (ß, - 0.09; 95%CI, - 0.13 to - 0.05; P = 1.21 × 10-5). CONCLUSIONS: Sleep habits not only result from but also contribute to alterations in brain structure, which may shed light on the possible mechanisms linking sleep behaviours with neuropsychiatric disorders, and offer new strategies for prevention and intervention in psychiatric disorders and sleep disturbance.