Search Results (897)

Accurate spleen segmentation in abdominal CT scans remains a critical challenge in medical image analysis due to variable morphology, low tissue contrast, and proximity to similar anatomical structures. This paper presents an enhanced U-Net architecture that addresses these challenges through multi-slice contextual integration and interpretable deep learning. Our approach incorporates three-channel inputs from adjacent CT slices, implements a hybrid loss function combining Dice and binary cross-entropy terms, and integrates Grad-CAM visualization for enhanced model interpretability. Comprehensive evaluation on the Medical Decathlon dataset demonstrates superior performance, with a Dice similarity coefficient of 0.923 ± 0.04, outperforming standard 2D approaches by 3.2%. The model exhibits robust performance across varying slice thicknesses, contrast phases, and pathological conditions. Grad-CAM analysis reveals focused attention on spleen–tissue interfaces and internal vascular structures, providing clinical insight into model decision-making. The system demonstrates practical applicability for automated splenic volumetry, trauma assessment, and surgical planning, with processing times suitable for clinical workflow integration. Full article

Background: This study aimed to develop and evaluate machine learning (ML) models to predict biochemical recurrence (BCR) after robot-assisted radical prostatectomy (RARP). Methods: We retrospectively analyzed clinical data from 1125 patients who underwent RARP between July 2013 and December 2023. The dataset was divided into a training set (70%) and a testing set (30%) using a stratified sampling strategy. Five ML models were developed using the training set. Model performance was evaluated on the testing set using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, and F1 scores. Additionally, model interpretability was assessed using SHapley Additive exPlanations (SHAP) values to determine the contribution of individual features. Results: Among the five ML models, the LightGBM model achieved the best prediction ability with an AUC of 0.881 (95%CI: 0.840–0.922) in the testing set. For model interpretability, SHAP values explained the contribution of individual features to the model, revealing that pathological T stage (pT), positive surgical margin (PSM), prostate-specific antigen (PSA) nadir, initial PSA, systematic prostate biopsy positive rate, seminal vesicle invasion (SVI), pathological International Society of Urological Pathology Grade Group (pGG), and perineural invasion (PI) were the key contributors to the predictive performance. Conclusions: We developed and validated ML models to predict BCR following RARP and identified that the LightGBM model with 8 variables achieved promising performance and demonstrated a high level of clinical applicability. Full article

The completeness and integrity of multimodal medical data are critical determinants of surgical success and postoperative recovery. However, because of issues such as poor sensor contact, small vibrations, and device discrepancies during signal acquisition, there are frequent missing values in patients’ medical data. This issue is especially prominent in rare or complex cases, where the inherent complexity and sparsity of multimodal data limit dataset diversity and degrade predictive model performance. As a result, clinicians’ understanding of patient conditions is restricted, and the development of robust algorithms to predict preoperative, intraoperative, and postoperative disease progression is hindered. To address these challenges, we propose Med-Diffusion, a diffusion-based generative framework designed to enhance sensor data by imputing missing multimodal clinical data, including both categorical and numerical variables. The framework integrates one-hot encoding, simulated bit encoding, and feature tokenization to improve adaptability to heterogeneous data types, utilizing conditional diffusion modeling for accurate data completion. Med-Diffusion effectively learns the underlying distributions of multimodal datasets, synthesizing plausible data for incomplete records, and it mitigates the data sparsity caused by poor sensor contact, vibrations, and device discrepancies. Extensive experiments demonstrate that Med-Diffusion accurately reconstructs missing multimodal clinical information and significantly enhances the performance of downstream predictive models. Full article

Patient-specific technologies within the field of adult spinal deformity (ASD) aid surgeons in pre-surgical planning, accurately help identify anatomical landmarks, and can project optimal post-surgical sagittal alignment. This narrative review aims to discuss the current uses of patient-specific technologies in ASD and identify new innovations that may very soon be integrated into patient care. Pre-operatively, machine learning or artificial intelligence helps surgeons to simulate post-operative alignment and provide information for the 3D-printing of pre-contoured rods and patient-specific cages. Intraoperatively, robotic surgery and intraoperative guides allow for more accurate positioning of implants. Implant materials are being developed to allow for better osseointegration and patient outcome monitoring. Despite the significant promise of these technologies, work still needs to be performed to ensure their accuracy, safety, and cost efficacy. Full article

Background/Objectives: Cesarean section (C-section) is a common surgical procedure associated with an increased risk of 30-day postpartum hospital readmissions. This study utilized machine learning (ML) to predict readmissions using a nationwide database. Methods: A retrospective analysis of the National Surgical Quality Improvement Project (2012–2022) included 54,593 patients who underwent C-sections. Random Forests (RF) and Extreme Gradient Boosting (XGBoost) models were developed and compared to logistic regression (LR) using demographic, preoperative, and perioperative data. Results: Of the cohort, 1306 (2.39%) patients were readmitted. Readmitted patients had higher rates of being of African American race (17.99% vs. 9.83%), diabetes (11.03% vs. 8.19%), and hypertension (11.49% vs. 4.68%) (p < 0.001). RF achieved the highest performance (AUC = 0.737, sensitivity = 72.03%, specificity: 61.33%), and a preoperative-only RF model achieved a sensitivity of 83.14%. Key predictors included age, BMI, operative time, white blood cell count, and hematocrit. Conclusions: ML effectively predicts C-section readmissions, supporting early identification and interventions to improve patient outcomes and reduce healthcare costs. Full article

Robot-assisted radical prostatectomy (RARP) has become the most prevalent treatment for patients with organ-confined prostate cancer. Despite superior outcomes, suboptimal vesicourethral anastomosis (VUA) may lead to serious complications, including urinary leakage, prolonged catheterization, and extended hospitalization. A precise localization of both the surgical needle and the surrounding vesical and urethral tissues to coadapt is needed for fine-grained assessment of this task. Nonetheless, the identification of anatomical structures from endoscopic videos is difficult due to tissue distortions, changes in brightness, and instrument interferences. In this paper, we propose and compare two Deep Learning (DL) pipelines for the automatic segmentation of the mucosal layers and the suturing needle in real RARP videos by exploiting different architectures and training strategies. To train the models, we introduce a novel, annotated dataset collected from four VUA procedures. Experimental results show that the nnU-Net 2D model achieved the highest class-specific metrics, with a Dice Score of 0.663 for the mucosa class and 0.866 for the needle class, outperforming both transformer-based and baseline convolutional approaches on external validation video sequences. This work paves the way for computer-assisted tools that can objectively evaluate surgical performance during the critical phase of suturing tasks. Full article

Breast cancer is one of the major health threats to women worldwide; thus, a need has arisen to reduce the number of instances and deaths through new methods of diagnostic monitoring and treatment. The present review is the synthesis of the recent clinical studies and technological advances in the application of magnetic resonance imaging (MRI) to monitor the pharmacological treatment of breast cancer. The specific focus is on high-risk groups (carriers of BRCA mutations and recipients of neoadjuvant chemotherapy) and the use of novel MRI methods (dynamic contrast-enhanced (DCE) MRI, diffusion-weighted imaging (DWI), and radiomics tools). All the reviewed studies show that MRI is more sensitive (up to 95%) and specific than conventional imaging in detecting malignancy particularly in dense breast tissue. Moreover, MRI can be used to assess the response and residual disease in a tumor early and accurately for personalized treatment, de-escalate unneeded interventions, and maximize positive outcomes. AI-based radiomics combined with deep-learning models also expand the ability to predict the therapeutic response and molecular subtypes, and can mitigate the risk of overfitting models when using complex methods of modeling. Other developments are hybrid PET/MRI, image guidance during surgery, margin assessment intraoperatively, three-dimensional surgical templates, and the utilization of MRI in surgery planning and reducing reoperation. Although economic factors will always play a role, the diagnostic and prognostic accuracy and capability to aid in targeted treatment makes MRI a key tool for modern breast cancer. The growing complement of MRI and novel curative approaches indicate that breast cancer patients may experience better survival and recuperation, fewer recurrences, and a better quality of life. Full article

Background: Uniportal video-thoracoscopic lobectomy has improved postoperative outcomes in lung cancer patients. Thus, thoracic surgeons are increasingly required to learn this new approach. Methods: We evaluate the path of a single surgeon switching from triportal video-thoracoscopic lobectomy to the uniportal, using the cumulative sum (CUSUM) analysis, in a single center to assess the learning curve, enrolling 107 uniportal video-thoracoscopic lobectomies consecutively performed. CUSUM analysis detected how many uniportal video-thoracoscopies occur to obtain changes in mean operation time, among all procedures consecutively performed. CUSUM analysis identified the cut-off at the 67th procedure; this value was used to divide all patients into two groups: group A (first 67 patients, early phase) and group B (40 patients, experienced phase). Then, we analyze the perioperative outcomes between the two groups. Results: Gender characteristics of the two groups were statistically similar. Median operative time decreased significantly after the early phase [188 min (IQR: 151–236) vs. 170.5 (IQR: 134–202) (p-value = 0.02)], respectively. Similarly, during the second phase, the conversions rate decreased: [10 (15%) (group A) vs. 1 (2%) (group B) (p-value = 0.04)], as did the postoperative complications [28 cases (42%) vs. 9 cases (22%) (p-value = 0.04)] and the length of stay [6 days (IQR 5–9.5) vs. 5 days (IQR 4–8) (p-value = 0.04)], giving evidence of skills acquired in the second phase. Conclusions: CUSUM analysis identified 67 uniportal lobectomies, after which operative time, conversion rate, and perioperative complications significantly decreased; the moving average analysis further supports a progressive reduction in operative time. Despite prior multiportal video-thoracoscopic experience, switching to uniportal video-thoracoscopy requires a distinct learning process. Full article

Background/Objectives: Surgical ward rounds are central to trainee education but are often associated with stress, cognitive overload, and inconsistent learning. Advances in artificial intelligence (AI), particularly large language models (LLMs), offer new ways to support trainees by simulating ward-round questioning, enhancing preparedness, and reducing anxiety. This study explores the role of generative AI in surgical ward-round education. Methods: Hypothetical plastic and reconstructive surgery ward-round scenarios were developed, including flexor tenosynovitis, DIEP flap monitoring, acute burns, and abscess management. Using de-identified vignettes, AI platforms (ChatGPT-4.5 and Gemini 2.0) generated consultant-level questions and structured responses. Outputs were assessed qualitatively for relevance, educational value, and alignment with surgical competencies. Results: ChatGPT-4.5 showed a strong ability to anticipate consultant-style questions and deliver concise, accurate answers across multiple surgical domains. ChatGPT-4.5 consistently outperformed Gemini 2.0 across all domains, with higher expert Likert ratings for accuracy, clarity, and educational value. It was particularly effective in pre-ward round preparation, enabling simulated questioning that mirrored consultant expectations. AI also aided post-round consolidation by providing tailored summaries and revision materials. Limitations included occasional inaccuracies, risk of over-reliance, and privacy considerations. Conclusions: Generative AI, particularly ChatGPT-4.5, shows promise as a supplementary tool in surgical ward-round education. While both models demonstrated utility, ChatGPT-4.5 was superior in replicating consultant-level questioning and providing structured responses. Pilot programs with ethical oversight are needed to evaluate their impact on trainee confidence, performance, and outcomes. Although plastic surgery cases were used for proof of concept, the findings are relevant to surgical education across subspecialties. Full article

Evidence of possible cerebellar involvement in spatial processing, place learning and other types of higher order functions comes mainly from clinical observations, as well as from mutant mice and lesion studies. The latter, in particular, have reported deficits in spatial learning and memory following surgical or neurotoxic cerebellar ablation. However, the low specificity of such manipulations has often made it difficult to precisely dissect the cognitive components of the observed behaviors. Likewise, due to conflicting data coming from lesion studies, it has not been possible so far to conclusively address whether a cerebellar dysfunction is sufficient per se to induce learning deficits, or whether concurrent damage to other regulatory structure(s) is necessary to significantly interfere with cognitive processing. In the present study, the immunotoxin 192 IgG-saporin, selectively targeting cholinergic neurons in the basal forebrain and a subpopulation of cerebellar Purkinje cells, was administered to adult rats bilaterally into the basal forebrain nuclei, the cerebellar cortices or both areas combined. Additional animals underwent injections of the toxin into the lateral ventricles. Starting from two–three weeks post-lesion, the animals were tested on paradigms of motor ability as well as spatial learning and memory and then sacrificed for post-mortem morphological analyses. All lesioned rats showed no signs of ataxia and no motor deficits that could impair their performance in the water maze task. The rats with discrete cerebellar lesions exhibited fairly normal performance and did not differ from controls in any aspect of the task. By contrast, animals with double lesions, as well as those with 192 IgG-saporin given intraventricularly did manifest severe impairments in both reference and working memory. Histo- and immunohistochemical analyses confirmed the effects of the toxin conjugate on target neurons and fairly similar patterns of Purkinje cell loss in the animals with cerebellar lesion only, basal forebrain-cerebellar double lesions and bilateral intraventricular injections of the toxin. No such loss was by contrast seen in the basal forebrain-lesioned animals, whose Purkinje cells were largely spared and exhibited a normal distribution pattern. The results suggest important functional interactions between the ascending regulatory inputs from the cerebellum and those arising in the basal forebrain nuclei that would act together to modulate the complex sensory–motor and cognitive processes required to control whole body movement in space. Full article

Sedation is an essential component of the anesthesia process. Inadequate sedation during anesthesia increases the risk of patient discomfort, intraoperative awareness, and psychological trauma. Conventional electroencephalography (EEG) based depth of anesthesia monitoring is often impractical in out-of-hospital settings due to equipment limitations and signal artifacts. Alternative non-EEG-based approaches are therefore required. In this study, we developed a machine learning model to detect inadequate sedation using 27 feature parameters, including demographics, vital signs, and heart rate variability metrics, from the open-access VitalDB database. Patient states were defined as inadequate sedation when the bispectral index (BIS) > 60. We systematically evaluated four temporal windows and four algorithms, and assessed model interpretability using Shapley Additive Explanations (SHAP). The Light Gradient Boosting Machine (LGBM) achieved the best performance, with an area under the receiver operating characteristic curve (AUC) of 0.825 and an accuracy (ACC) of 0.741 using a 2 s time window. Extending the time window to 20 s improved both metrics by approximately 0.012. Feature selection identified 12 key parameters that maintained comparable accuracy, confirming robustness with reduced complexity. These findings demonstrate the feasibility of using non-EEG-based physiological data for real-time detection of inadequate sedation. The developed model is interpretable, resource-efficient, scalable, and shows strong potential for integration into portable monitoring systems in prehospital, emergency, and low-resource surgical settings. Full article

Artificial intelligence (AI) and augmented intelligence have significant potential in oculoplastics, offering tools for diagnosis, treatment recommendations, and administrative efficiency. This article discusses current and potential applications of AI in ptosis, eyelid and conjunctival cancer, thyroid-associated orbitopathy (TAO), giant cell arteritis (GCA), and orbital fractures. AI-based programs can assist in screening, predicting surgical outcomes, and improving patient care through data-driven decisions. Privacy concerns, particularly with the use of facial and ocular photographs, require robust solutions, including blockchain, federated learning and steganography. Large generalizable datasets with adequate validation are crucial for future AI development. While AI can assist in clinical decision-making and administrative tasks, physician oversight remains critical to prevent potential errors. Large language models like ChatGPT also have the potential to counsel patients, although further validation is needed to ensure accuracy and patient safety. Ultimately, AI should be regarded as an augmentative tool that supports, rather than replaces, physician expertise in oculoplastic care. Full article

Background/Objectives: Escherichia coli is one of the leading causes of surgical site infections (SSIs) and poses a growing public health concern due to its increasing antimicrobial resistance. High rates of extended-spectrum beta-lactamase (ESBL) production among E. coli strains complicate treatment outcomes and emphasize the need for effective surveillance and control strategies. Methods: A total of 691 E. coli isolates from general surgery clinics (2020–2025) were identified using MALDI-TOF MS. Antibiotic susceptibility data and patient variables were cleaned, encoded, and used to predict resistance using the Random Forest, CatBoost, and Naive Bayes algorithms. SMOTE addressed class imbalance, and model performance was assessed through various validation methods. Results: Among the three machine learning models tested, Random Forest (RF) showed the best performance in predicting antibiotic resistance of E. coli, achieving median accuracy, precision, recall, and F1-scores of 0.90 and AUC values up to 0.99 for key antibiotics. CatBoost performed similarly but was less stable with imbalanced data, while Naive Bayes showed lower accuracy. Feature importance analysis highlighted strong inter-antibiotic resistance links, especially among β-lactams, and some influence of demographic factors. Conclusions: This study highlights the potential of simple, high-performing models using structured clinical data to predict antimicrobial resistance, especially in resource-limited clinical settings. By incorporating machine learning into antimicrobial resistance (AMR) surveillance systems, our goal is to support the advancement of rapid diagnostics and targeted antimicrobial stewardship approaches, which are essential in addressing the growing challenge of multidrug resistance. Full article

Intraoperative radiotherapy (IORT) is a radiation technique that allows for the delivery of a high radiation dose to the target while preserving the surrounding structures, which can be displaced during the surgical procedure. An important limitation of this technique is the lack of real-time image guidance, which is one of the main achievements of modern radiation therapy because it allows for treatment optimization. IORT can be delivered by low-energy X-rays or by accelerated electrons. The present review describes the most relevant clinical applications for IORT and discusses the potential advantages of using artificial intelligence (AI) to overcome some of the current limitations of IORT. In recent decades, IORT has proven to be an effective treatment in several cancer types. In breast cancer, IORT can be used to deliver a single dose of radiation (partial breast irradiation) or as a boost in high-risk patients. In locally advanced rectal cancer, a single dose to the tumor bed can improve local control and prevent pelvic relapse in primary and recurrent tumors. In sarcomas, IORT enables the delivery of high doses, achieving good functional outcomes with low toxicity in tumors located in the retroperitoneum and extremities. In pancreatic cancer, IORT shows promising results in borderline resectable and unresectable cases. Ongoing technological advances are addressing current challenges in imaging and radiation planning, paving the way for personalized, image-guided IORT. Recent innovations such as CT- and MRI-equipped hybrid operating theaters allow for real-time imaging, which could be used for AI-assisted segmentation and planning. Moreover, the implementation of AI in terms of machine learning, deep learning, and radiomics can improve the interpretation of imaging, predict treatment outcomes, and optimize workflow efficiency. Full article

Objective: The role of the frozen elephant trunk technique in the treatment of acute aortic dissections is currently based on results from high-volume centers only. We investigated the patient selection process, intraoperative data, the evolution of surgical practice and outcomes from a low-volume center. Methods: A retrospective analysis was conducted on 202 acute aortic dissection (AAD) patients treated between October 2014 and December 2023. Patients were categorized into those receiving less invasive open aortic repair (group 1, n = 136) and those undergoing frozen elephant trunk procedures (FETs) (group 2, n = 66). Data on demographics, surgical procedures, and outcomes were analyzed. Results: Overall 30-day mortality was 16% (13% vs. 23%; p = 0.068). Rates of postoperative disabling stroke were similar (9% vs. 8%, p = 0.190). FET procedures required longer cardiopulmonary bypass (195 min vs. 234 min, p = 0.011), hypothermic circulatory arrest (26 min vs. 43 min, p < 0.001), and selective cerebral perfusion times (26 min vs. 47 min, p < 0.001). Follow-up indicated that 17% of FET patients received completion thoracic endovascular aortic repair (TEVAR) versus 4% in non-FET patients (p = 0.002), whereas no difference was seen in open surgical reintervention. Median follow-up at 33 months showed an overall mortality of 27%, with no significant difference between groups (23% in group 1 vs. 35% in group 2, p = 0.123). Conclusions: The FET technique is feasible in low-volume centers, yielding outcomes comparable to high-volume centers. FET proximalization and a liberal use of extra-anatomical left subclavian artery (LSA) grafts ease the learning curve. Completion treatments can be effectively conducted following FET implantation to further induce positive aortic remodelling. Full article