Mapping surgery is fundamental to developing operative guidelines and enabling autonomous robotic surgery. Recent advances in artificial intelligence (AI) have shown promise in mapping the behaviour of surgeons from videos, yet current models remain narrow in scope, capturing limited behavioural components within single procedures, and offer limited translational value, as they remain inaccessible to practising surgeons. Here we introduce Halsted, a vision-language model trained on the Halsted Surgical Atlas (HSA), one of the most comprehensive annotated video libraries grown through an iterative self-labelling framework and encompassing over 650,000 videos across eight surgical specialties. To facilitate benchmarking, we publicly release HSA-27k, a subset of the Halsted Surgical Atlas. Halsted surpasses previous state-of-the-art models in mapping surgical activity while offering greater comprehensiveness and computational efficiency. To bridge the longstanding translational gap of surgical AI, we develop the Halsted web platform (https://halstedhealth.ai/) to provide surgeons anywhere in the world with the previously-unavailable capability of automatically mapping their own procedures within minutes. By standardizing unstructured surgical video data and making these capabilities directly accessible to surgeons, our work brings surgical AI closer to clinical deployment and helps pave the way toward autonomous robotic surgery.
Machine unlearning (MU) aims to efficiently remove sensitive or harmful memory from a pre-trained model. The key challenge is to balance the potential tradeoff between unlearning efficacy and utility preservation, which involves forgetting undesirable information as defined while maintaining the model's original performance. One potential way to tackle this problem is to use multi-objective optimization to jointly optimize both the unlearning and utility preservation objectives. However, existing multi-objective methods only guarantee finding a Pareto-optimal solution without fine-grained control, which causes under-optimization of the unlearning objective. To this end, we first model MU as a constrained optimization problem, that is, optimizing the unlearning objective under the constraint of a bounded increase for utility loss. We then show that solving this optimization problem is equivalent to unilateral gradient surgery on the unlearning objective. To resolve the additional computational cost brought by gradient surgery, we propose an implicit gradient surgery method, which approximates the solution to the aforementioned constrained optimization problem via only one backpropagation, thereby achieving efficient utility-preserving MU. Theoretically, we provide a tight convergence analysis of the algorithm. Empirically, our extensive experiments show that the proposed algorithm achieves better tradeoff results than existing baselines. Codes are available at https://github.com/anseryuer/EUPMU-Efficient-Utility-Preserving-Machine-Unlearning.
We establish a $d$-invariant surgery formula for $L$-space knots that provides an effective tool for studying surgeries between lens spaces. Using this formula, we classify distance one surgeries between lens spaces of the form $L(n,1)$. This classification has direct applications to band surgeries between torus links $T(2,n)$, with connections to DNA topology. In particular, we show that chirally cosmetic banding of torus links can possibly occur only when $n=1,5,9$ or $10$.
The $\hat{Z}$ invariants of three-manifolds introduced by Gukov-Pei-Putrov-Vafa have influenced many areas of mathematics and physics. However, their TQFT structure remains poorly understood. In this work, we develop a framework of decorated $\mathrm{Spin}$-TQFTs and construct one based on Atiyah-Segal-like axioms that computes the $\hat{Z}$ invariants. Central to our approach is a novel quantization of $SL(2,\mathbb{C})$ Chern-Simons theory and a $\mathbb{Q}$-extension of the algebra of observables on the torus, from which we obtain the torus state space of the $\hat{Z}$-TQFT. Using the torus state space and topological invariance, we uniquely determine the $\hat{Z}$ invariants for negative-definite plumbed manifolds. Within this TQFT framework, we establish gluing, rational surgery, partial surgery, satellite, and cabling formulas, as well as explicit closed-form expressions for Seifert manifolds and torus link complements. We also generalize these constructions to higher-rank gauge groups.
Precise segmentation of individual teeth from digital three-dimensional (3D) tooth models is critical in computer-assisted orthodontic surgery. This study explores the application of Point Multi-Layer Perceptron (PointMLP) in processing 3D tooth models and introduces an innovative integration of the Graph Attentional Convolution (GAC) Layer with a graph attention mechanism. By incorporating the GAC Layer into PointMLP, the model can focus on key local regions in the 3D tooth model and dynamically adjust the attention applied to these areas. This enhanced attention mechanism allows the model to better capture subtle surface structures, facilitating the accurate extraction of valuable local features. Compared to other traditional segmentation algorithms, the proposed method shows improvements of 1.1, 2.04, 1.06, 2.2, and 1.8 percentage points in Overall Accuracy (OA), Sensitivity (SEN), Positive Predictive Value (PPV), and Intersection Over Union (IoU), respectively. At the same number of training epochs, our method outperforms both GAC and PointMLP in segmentation performance.
Control engineering systems. Automatic machinery (General), Systems engineering
We prove that for any pair of Legendrian representatives of the Chekanov-Eliashberg twist knots with different LOSS invariants, any negative rational contact $r$-surgery with $r\neq -1$ always gives rise to different contact 3-manifolds distinguished by their contact invariants. This gives the first examples of pairs of Legendrian knots with the same classical invariants but distinct contact $r$-surgeries for all negative rational number $r$. We also generalize the statement from the twist knots to a certain families of two-bridge knots.
This paper is a comprehensive exploring of technology capability in 5G/6G TIS, explicitly focusing on the potential of remote surgery globally and in Germany. The paper's main contribution is its ability to anticipate new debates on the interplay between TIS and contexts, with particular emphasis on the national and international levels. Our findings, derived from a Bibliometrics study of industry-academic relationships, highlight crucial collaborations in Germany, positioning the country as a strategic actor in international TIS and, by extension, in applying 5G/6G technological systems to remote surgery due to its knowledge production capability. We propose policies that can stimulate interaction between smaller suppliers and larger companies, which can act as intermediaries and provide access to international markets.
Mahmoud Kassem, Ahmed Kamr, Creighton B. Wright
et al.
Objective: Breast conserving surgery is an excellent option in the treatment of breast cancer. To achieve a good result with this modality, a surgeon needs to identify and excise the tumor with adequate margins. The radiofrequency identification (RFID) technique is a wireless localization technique used for intraoperative breast lesion identification. We assessed the efficacy and outcomes of the RFID technique in breast cancer patients at our institution. Materials and Methods: This is a single institution, retrospective study (BSMH 22-02X-MWH) of 73 patients. We analyzed the medical records of women with biopsy-proven breast cancer from June 2020 to August 2022; participants received surgical care at Mercy Health West Hospital. Data collected included demographics, clinicopathological characteristics, and surgical procedure. The primary objective was to determine the safety and efficacy of RFID. The secondary objective was to assess the impact of obesity and breast density on the RFID outcomes. Results: A total of 73 female patients met the eligibility criteria with stage I (59%) and grade 1 (51%) breast cancer with mean age of at diagnosis of 66.8 years and mean body mass index of 31.4 kg/m2. Patients had invasive ductal carcinoma (61%), hormonal positive (56%), and human epidermal growth factor receptor 2 negative (68%) disease. All RFID tags were placed under image guidance with 100% accuracy of placement with no evidence of migration or procedure revision. Ninety percent of patients had free surgical margins and only seven patients needed margin re-excision with successful removal of the lesion and the tag. Conclusion: RFID localization technique is a safe, effective and reliable procedure that results in favorable patient outcomes and quality of life.
Neoplasms. Tumors. Oncology. Including cancer and carcinogens, Medicine
Wenjie Cai, Yvonne Forsell, Catharina Lavebratt
et al.
Abstract Associations between the fat mass and obesity-associated (FTO) gene and obesity are well-established. However, recent studies have linked FTO to addiction phenotypes and dopaminergic signaling, thus suggesting broader psychiatric implications. We explored this assumption by conducting a phenome-wide association study across 4756 genome-wide association studies, identifying 23–26 psychiatric traits associated with FTO at the multiple-corrected significance level. These traits clustered into four categories: substance use, chronotype/sleep, well-being, and neuroticism. To validate these findings, we analyzed a functionally suggestive FTO variant (rs1421085) in a separate cohort, examining its impact on (i) alcohol use based on the Alcohol Use Disorders Identification Test (AUDIT), (ii) subjective well-being based on the WHO (Ten) Well-Being Index, and (iii) neuroticism based on Schafer’s Five Factor Model or the Karolinska Scales of Personality. Our results confirmed a direct association between rs1421085 and neuroticism that was independent of age, sex, alcohol use, body mass index (BMI), and childhood adversities. Interestingly, while no direct association with alcohol intake was observed, both cross-sectional and lagged longitudinal mediation analyses uncovered indirect relationships between rs1421085 and problematic alcohol use (AUDIT-P), with increased neuroticism acting as the intermediary. Mediation analyses also supported an indirect effect of rs1421085 on lower well-being through the pathways of increased neuroticism and BMI. Our study is the first to validate a direct association between FTO and neuroticism. However, additional studies are warranted to affirm the causal pathways linking FTO to well-being and alcohol use through neuroticism.