Citations

Abstract Breast cancer is one of the most common cancers in women. With early diagnosis, some breast cancers are highly curable. Classifying normal versus tumor breast tissues from microscopy images is an ideal case to use for deep learning and could help diagnose breast cancer with higher reproducibility. Since data preprocessing and hyperparameter tuning have impacts on breast cancer classification accuracies, training a classifier with appropriate data preprocessing and optimized hyperparameters could improve breast cancer classification accuracy. Using 12 combinations of model architectures, data preprocessing, and hyperparameter configurations, the validation accuracy was calculated using the BreAst Cancer Histology (BACH) dataset. The DenseNet201, a non-specialized model architecture, with transfer learning approach achieved 98.61% validation accuracy compared to only 64.00% for the digital pathology (DP)-specialized model architecture. The combination of image data preprocessing and hyperparameters have a profound impact on the performance of deep neural networks for image classification. To identify a well-performing neural network to classify tumor versus normal breast histology, researchers should not only focus on developing new models specifically for DP, since hyperparameter tuning for existing deep neural networks in the computer vision field could also achieve a better prediction accuracy.

Background: Targeting interleukin-23 (IL-23) represents a significant therapeutic avenue for treating ulcerative colitis (UC). Study Question: What are the effectiveness and safety of selective inhibitors targeting IL-23p19 and IL-12/23p40 in individuals with moderate-to-severe UC? Data Sources: MEDLINE, Embase, Scopus, and Cochrane databases. Study Design: A systematic search of MEDLINE, Embase, Scopus, and Cochrane databases till January 15, 2024, to identify randomized controlled trials comparing IL-23p19 and IL-12/23p40 inhibitors against placebo or active comparators in UC patients. The primary outcome was clinical remission, with secondary outcomes including clinical response, endoscopic remission, and safety profiles during induction and maintenance phases. Using a fixed-effect model, we pooled dichotomous data with risk ratio (RR) and 95% confidence interval (CI) for analysis. Results: In 5 trials involving 1120 patients with moderate to severe UC, targeting IL-23 showed significant superiority in inducing clinical remission [RR: 2.08, 95% CI, (1.66–2.61)], endoscopic remission [RR: 1.73, 95% CI, (1.39–2.16)], and histologic remission [RR: 1.88, 95% CI, (1.34–2.64)]. Additionally, individuals treated with IL-12/23p40 or IL-23p19 antagonists maintained clinical remission [RR: 1.85, 95% CI, (1.53–2.23)], endoscopic remission [RR: 2.03, 95% CI, (1.60–2.57)], and histologic remission [RR: 1.66, 95% CI, (1.11–2.48)]. Targeting IL-23 was linked with a reduced risk of any adverse events (AE) during both induction [RR: 0.94, 95% CI, (0.86–1.02)] and maintenance phases [RR: 0.93, 95% CI, (0.86–0.99)], any serious AE during the induction phase [RR: 0.53, 95% CI, (0.36–0.78)], and withdrawal due to AEs compared to patients receiving placebo during induction [RR: 0.24, 95% CI (0.14, 0.43)]. Conclusion: Targeting IL-23 demonstrates efficacy and safety for inducing and maintaining clinical and endoscopic remission in moderate-to-severe UC patients.

Background Although flow diverters (FDs) have benefited from several technical improvements, recently concerns have arisen regarding the braid stability after implantation. Thus, we investigated frequency, predictive factors, and clinical impact of the phenomenon of FD braid deformation (FDBD). Methods Consecutive intracranial aneurysms (IAs) treated with various FDs, between January 2018 and July 2023, were reviewed to identify FDBD (defined as the deformation of a FD without any external force applied to it). Patient, aneurysm, procedural, and FD characteristics were retrieved and analyzed using univariate and multivariable analyses. Morbidity is defined as a score of +1 in the modified Rankin Scale at 3 months. Results In total, 245 FD procedures (271 FDs implanted; 25 multiple IAs treated with 1 FD) in 228 patients; FDBD was observed in 36/245 cases (14.7%), mainly at follow-up angiography (32/36, 88.9%); fish-mouthing was the most frequent FDBD. Morbidity was related to fish-mouthing and braid collapse and was significantly higher in the FDBD group after retreatment (p=0.04). Drawn filled tubing with platinum (DFT) (adjusted odds ratio (aOR)=7.0, 95% CI 3.0 to 17.5; p<0.001) and FD diameter (aOR=2.2, 95% CI 1.3 to 4.1; p<0.01) were identified as independent predictors of FDBD. The metal alloy composing the FD (p=0.13) and coated surfaces were not significantly associated with FDBD (p=0.54 in multivariable analysis). Conclusions FDBD is a frequent phenomenon observed in about 15% of cases, and it was responsible for higher morbidity. Only FD characteristics (DFT and FD diameter) were independent determinants of FDBD. Future research should focus on the impact of novel braid configurations and materials on braid stability.

DddA-derived cytosine base editors (DdCBEs) enable the targeted introduction of C•G-to-T•A conversions in mitochondrial DNA (mtDNA). DdCBEs work in pairs, with each arm composed of a transcription activator-like effector (TALE), a split double-stranded DNA deaminase half, and a uracil glycosylase inhibitor. This pioneering technology has helped improve our understanding of cellular processes involving mtDNA and has paved the way for the development of models and therapies for genetic disorders caused by pathogenic mtDNA variants. Nonetheless, given the intrinsic properties of TALE proteins, several target sites in human mtDNA are predicted to remain out of reach to DdCBEs and other TALE-based technologies. Specifically, due to the conventional requirement for a thymine immediately upstream of the TALE target sequences ( i.e. , the 5′-T constraint), over 150 loci in the human mitochondrial genome are presumed to be inaccessible to DdCBEs. Previous attempts at circumventing this requirement, either by developing monomeric DdCBEs or utilizing DNA-binding domains alternative to TALEs, have resulted in suboptimal specificity profiles with reduced therapeutic potential. Here, aiming to challenge and elucidate the relevance of the 5′-T constraint in the context of DdCBE-mediated mtDNA editing, and to expand the range of motifs that are editable by this technology, we generated DdCBEs containing TALE proteins engineered to recognize all 5′ bases. These modified DdCBEs are herein referred to as αDdCBEs. Notably, 5′-T-noncompliant canonical DdCBEs efficiently edited mtDNA at diverse loci. However, they were frequently outperformed by αDdCBEs, which exhibited significant improvements in activity and specificity, regardless of the most 5′ bases of their TALE binding sites. Furthermore, we showed that αDdCBEs are compatible with the enhanced DddA tox variants DddA6 and DddA11, and we validated TALE shifting with αDdCBEs as an effective approach to optimize base editing outcomes. Overall, αDdCBEs enable efficient, specific, and unconstrained mitochondrial base editing.

The AI-MIRACLE Study investigates the efficacy of using ChatGPT 4.0, a large language model (LLM), for translating and simplifying radiology reports into multiple languages, aimed at enhancing patient comprehension. The study assesses the model’s performance across the most spoken languages in the U.S., emphasizing the accuracy and clarity of translated and simplified radiology reports for non-medical readers. This study employed ChatGPT 4.0 to translate and simplify selected radiology reports into Vietnamese, Tagalog, Spanish, Mandarin, and Arabic. Hindi was used as a preliminary test language for validation of the questionnaire. Performance was assessed via Google form surveys distributed to bilingual physicians, which assessed the translation accuracy and clarity of simplified texts provided by ChatGPT 4. Responses from 24 participants showed mixed results. The study underscores the model’s varying success across different languages, emphasizing both potential applications and limitations. ChatGPT 4.0 shows promise in breaking down language barriers in healthcare settings, enhancing patient comprehension of complex medical information. However, the performance is inconsistent across languages, indicating a need for further refinement and more inclusive training of AI models to handle diverse medical contexts and languages. The study highlights the role of LLMs in improving healthcare communication and patient comprehension, while indicating the need for continued advancements in AI technology, particularly in the translation of low-resource languages.

Background: Primary lymphedema is a chronic condition caused by a developmental abnormality of the lymphatic system, leading to its malfunction. Various surgical options, including physiologic and excisional procedures, have been proposed. The aim of this study was to present a comprehensive algorithm for the treatment of primary lower extremity lymphedema: the Primary LYmphedema Multidisciplinary Approach (P-LYMA). Methods: Nineteen patients were treated following the P-LYMA protocol. Patients underwent pre- and postoperative complex decongestive therapy (CDT). A variety of physiologic and excisional procedures were performed, either independently or in combination. The primary outcome was to assess the circumferential reduction rate (CRR). The Lymphedema Quality of Life Score (LeQOLiS), reduction in the number of cellulitis episodes, and complications were recorded. Results: The mean CRR was 73 ± 20% at twelve months postoperatively. The frequency of cellulitis episodes per year decreased from a mean of 1.9 ± 0.8 preoperatively to 0.4 ± 0.6 during follow-up. Two patients experienced minor complications. The mean hospitalization time was 5 days. Patients’ quality of life, as measured by the LeQOLiS, significantly improved from 70.4 ± 12 preoperatively to 24 ± 14 at twelve months postoperatively. Conclusions: The P-LYMA algorithm maximizes surgical outcomes and improves the quality of life in patients with primary lymphedema. CDT is essential for optimizing results.

Abstract Meningiomas are the most frequent primary intracranial tumors. Hence, they constitute a major share of diagnostic specimens in neuropathology practice. The 2021 WHO Classification of Central Nervous System Tumors (“CNS5”) has introduced the first molecular grading parameters for meningioma with oncogenic variants in the TERT promoter and homozygous deletion of CDKN2A/B as markers for CNS WHO grade 3. However, after the publication of the new classification volume, clarifications were requested, not only on novel but also on long-standing questions in meningioma grading that were beyond the scope of the WHO “blue book.” In addition, more recent research into possible new molecular grading parameters could not yet be implemented in the 2021 classification but constitutes a compelling body of literature. Hence, the consortium to inform molecular and practical approaches to CNS tumor taxonomy-not official WHO (cIMPACT-NOW) Steering Committee convened a working group to provide such clarification and assess the evidence of possible novel molecular criteria. As a result, this cIMPACT-NOW update provides guidance for more standardized morphological evaluation and interpretation, most prominently pertaining to brain invasion, identifies scenarios in which advanced molecular testing is recommended, proposes to assign CNS WHO grade 2 for cases with CNS WHO grade 1 morphology but chromosomal arm 1p deletion in combination with 22q deletion and/or NF2 oncogenic variants, and discusses areas in which the current evidence is not yet sufficient to result in new recommendations.