We present CuMPerLay, a novel differentiable vectorization layer that enables the integration of Cubical Multiparameter Persistence (CMP) into deep learning pipelines. While CMP presents a natural and powerful way to topologically work with images, its use is hindered by the complexity of multifiltration structures as well as the vectorization of CMP. In face of these challenges, we introduce a new algorithm for vectorizing MP homologies of cubical complexes. Our CuMPerLay decomposes the CMP into a combination of individual, learnable single-parameter persistence, where the bifiltration functions are jointly learned. Thanks to the differentiability, its robust topological feature vectors can be seamlessly used within state-of-the-art architectures such as Swin Transformers. We establish theoretical guarantees for the stability of our vectorization under generalized Wasserstein metrics. Our experiments on benchmark medical imaging and computer vision datasets show the benefit CuMPerLay on classification and segmentation performance, particularly in limited-data scenarios. Overall, CuMPerLay offers a promising direction for integrating global structural information into deep networks for structured image analysis.

Foundation models open up new possibilities for the use of AI in healthcare. However, even when pre-trained on health data, they still need to be fine-tuned for specific downstream tasks. Furthermore, although foundation models reduce the amount of training data required to achieve good performance, obtaining sufficient data is still a challenge. This is due, in part, to restrictions on sharing and aggregating data from different sources to protect patients' privacy. One possible solution to this is to fine-tune foundation models via federated learning across multiple participating clients (i.e., hospitals, clinics, etc.). In this work, we propose a new personalized federated fine-tuning method that learns orthogonal LoRA adapters to disentangle general and client-specific knowledge, enabling each client to fully exploit both their own data and the data of others. Our preliminary results on real-world federated medical imaging tasks demonstrate that our approach is competitive against current federated fine-tuning methods.

Deep learning relies heavily on data augmentation to mitigate limited data, especially in medical imaging. Recent multimodal learning integrates text and images for segmentation, known as referring or text-guided image segmentation. However, common augmentations like rotation and flipping disrupt spatial alignment between image and text, weakening performance. To address this, we propose an early fusion framework that combines text and visual features before augmentation, preserving spatial consistency. We also design a lightweight generator that projects text embeddings into visual space, bridging semantic gaps. Visualization of generated pseudo-images shows accurate region localization. Our method is evaluated on three medical imaging tasks and four segmentation frameworks, achieving state-of-the-art results. Code is publicly available on GitHub: https://github.com/11yxk/MedSeg_EarlyFusion.

Although deep neural networks have provided impressive gains in performance, these improvements often come at the cost of increased computational complexity and expense. In many cases, such as 3D volume or video classification tasks, not all slices or frames are necessary due to inherent redundancies. To address this issue, we propose a novel learnable subsampling framework that can be integrated into any neural network architecture. Subsampling, being a nondifferentiable operation, poses significant challenges for direct adaptation into deep learning models. While some works, have proposed solutions using the Gumbel-max trick to overcome the problem of non-differentiability, they fall short in a crucial aspect: they are only task-adaptive and not inputadaptive. Once the sampling mechanism is learned, it remains static and does not adjust to different inputs, making it unsuitable for real-world applications. To this end, we propose an attention-guided sampling module that adapts to inputs even during inference. This dynamic adaptation results in performance gains and reduces complexity in deep neural network models. We demonstrate the effectiveness of our method on 3D medical imaging datasets from MedMNIST3D as well as two ultrasound video datasets for classification tasks, one of them being a challenging in-house dataset collected under real-world clinical conditions.

Following atrial fibrillation ablation, it is challenging to distinguish patients who will remain arrhythmia-free from those at risk for recurrence. New explainable machine learning (xML) techniques allow for systematic assessment of arrhythmia recurrence risk following catheter ablation. We aim to develop an xML algorithm that predicts recurrence and reveals key risk factors to facilitate better follow-up strategy after an ablation procedure. We reconstructed pre-and post-ablation models of the left atrium (LA) from late gadolinium enhanced magnetic resonance (LGE-MRI) for 67 patients. Patient-specific features (LGE-based measurements of pre/post-ablation arrhythmogenic substrate, LA geometry metrics, computational simulation results, and clinical risk factors) trained a random forest classifier to predict recurrent arrhythmia. We calculated each risk factor's marginal contribution to model decision making via SHapley Additive exPlanations (SHAP). The classifier accurately predicts post-ablation arrhythmia recurrence (mean receiver operating characteristic [ROC] area under the curve [AUC]: 0.80 ± 0.04; mean precision-recall [PR] AUC: 0.82 ± 0.08). SHAP analysis reveals that of 89 features tested, the key population risk factors for recurrence are: large left atrium, low LGE-quantified post-ablation scar in the atrial floor region, and previous attempts at direct current cardioversion. We also examine patient-specific recurrence predictions, since xML allows us to understand why a particular individual can have large prediction weights for some categories without tipping the balance towards an incorrect prediction. Finally, we validate our model in a completely new, 15-patient retrospective holdout cohort (80% correct). Our SHAP-based explainable machine learning approach is a proof-of-concept clinical tool to explain arrhythmia recurrence risk in patients who underwent ablation by combining patient-specific clinical profiles and LGE-derived data.

Proper stratification of recurrence risk in breast cancer is crucial for guiding treatment decisions. This study aims to predict the recurrence risk of breast cancer patients using a multimodal deep learning model that integrates multiple sequence MRI imaging features with clinicopathologic characteristics. In this retrospective study, we enrolled 574 patients with non-metastatic invasive breast cancer from two Chinese institutions between September 2012 and July 2019. We developed a multimodal deep learning (MDL) model by constructing a multi-instance learning framework based on convolutional neural networks. We integrated imaging features from T2WI, DWI, and DCE-MRI sequences with clinicopathologic features for breast cancer recurrence risk stratification. Subsequently, the performance of the MDL model was evaluated using receiver operating characteristic (ROC) curves, the Hosmer-Lemeshow test, calibration curves, and decision curve analysis (DCA). Survival analysis was conducted with Kaplan-Meier survival curves to stratify breast cancer patients into high and low-recurrence risk groups. Time-dependent ROC curves were used to assess 3-year, 5-year, and 7-year recurrence-free survival (RFS) for breast cancer patients. Additionally, we performed differential and enrichment analyses on Oncotype DX genes. We correlated these genes with clinicopathologic features and deep-learning radiographic features using univariate Cox regression and Pearson correlation analysis. The MDL model demonstrated good performance in predicting breast cancer recurrence risk and accurately differentiated between high- and low-recurrence risk groups, with an AUC as high as 0.915 (95% CI 0.8448-0.9856). The C-index of prediction models was 0.803 in the testing cohort. The AUCs for 5-year and 7-year RFS were 0.936 (95% CI 0.876-0.997) and 0.956 (95% CI 0.902-1.000) in the validation cohort. In the testing cohort, these AUCs were 0.836 (95% CI 0.763-0.909) and 0.783 (95% CI 0.676-0.891). This study found a significant correlation between Oncotype DX gene expression, clinicopathologic features, and deep-learning radiographic features (p < 0.05). This study validated the robust predictive accuracy of the MDL model in identifying high- and low-risk groups for recurrence. The correlations identified between Oncotype DX genes, clinicopathologic features, and deep-learning radiographic features offer novel insights for future biomarker research in breast cancer.

To develop and validate a deep learning tool for the automatic segmentation of pancreatic solid neoplasms and to establish a radiomics model for diagnosing these solid neoplasms in MRI. This retrospective study employed a three-dimensional nnU-Net-based model trained in plain MRI from patients who underwent resection for pancreatic neoplasms. A radiomics model was developed for diagnosing pancreatic neoplasms based on automatic segmentation. The segmentation performance of the deep learning model was quantitatively evaluated using dice similarity coefficient (DSC). The performance of the radiomics model was assessed through receiver operating characteristic analysis. The study included 165 and 89 patients in the training and testing cohorts. The deep learning model achieved excellent automatic segmentation performance, with mean DSC values of 0.82 on T2WI and 0.91 on DWI in the training cohort, and 0.64 on T2WI and 0.70 on DWI in the testing cohort, respectively. For pancreatic lesions smaller than 2 cm, the DSC values were 0.74 on T2WI and 0.92 on DWI in the training cohort, and 0.51 on T2WI and 0.62 on DWI in the testing cohort. Nine radiomics signatures were selected based on ROIs obtained from the automatic segmentation. The radiomics diagnostic model exhibited favorable performance in distinguishing pancreatic ductal adenocarcinomas (PDACs) from neuroendocrine neoplasms and solid pseudopapillary neoplasms, with AUCs of 0.968 and 0.790 in the training and testing cohorts, respectively. The deep learning automatic segmentation tool accurately detected pancreatic neoplasms in MRI scans, with reasonable efficiency for tumors smaller than 2 cm. The radiomics diagnostic model demonstrated favorable performance in differentiating PDACs from neuroendocrine neoplasms and solid pseudopapillary neoplasms.

This study aims to investigate the association between the retinal artery to vein ratio (AVR) and body fat distribution, and to further evaluate the potential beneficial effects of optimized fat distribution on systemic vascular health. A total of 2,698 participants aged 18 to 80 from Lanxi cohort were enrolled. After applying inclusion and exclusion criteria, 2,045 participants were retained for the final analysis. Retinal vessel images were obtained through fundoscopy, and retinal AVR was automatically calculated using a clinically validated AI algorithm. Body fat was assessed by dual-energy x-ray absorptiometry (DXA). Adjusted multivariate linear regression models were used to identify the associations of retinal AVR with fat distribution. Retinal AVR was negatively associated with waist-hip ratio (WHR), android fat mass percentage, android to gynoid fat ratio, and trunk fat mass percentage. Similar trends were observed when fat distribution indicators were categorized into quartiles (P for trend < 0.05). When stratified by age, a similar significant association was observed in the 45-60 age group. Retinal AVR was associated with fat distribution, with particularly correlations observed in middle-aged populations and those with metabolic abnormalities. And these associations differ based on the location of fat depots, indicating that exercise-induced fat redistribution is associated with vascular health.

This study evaluates the CT-based volumetric sarcopenia index (SI) as a baseline prognostic factor for overall survival (OS) in 10,340 solid tumor patients (40% female). Automated body composition analysis was applied to internal baseline abdomen CTs and to thorax CTs. SI's prognostic value was assessed using multivariable Cox proportional hazards regression, accelerated failure time models, and gradient-boosted machine learning. External validation included 439 patients (40% female). Higher SI was associated with prolonged OS in the internal abdomen (HR 0.56, 95% CI 0.52-0.59; P < 0.001) and thorax cohorts (HR 0.40, 95% CI 0.37-0.43; P < 0.001), as well as in the external validation cohort (HR 0.56, 95% CI 0.41-0.79; P < 0.001). Machine learning models identified SI as the most important factor in survival prediction. Our results demonstrate SI's potential as a fully automated body composition feature for standard oncologic workflows.

The selection of the appropriate size of a double-lumen tube (DLT) is a critical yet often underestimated aspect of thoracic anaesthesia. The present narrative review evaluates traditional and emerging methods for determining DLT size, including anthropometric formulas, chest X-rays, CT scans, and ultrasonography. Despite the prevalence of height- and gender-based predictions, mounting evidence underscores their restricted correlation with airway anatomy. Chest X-rays and CT scans have been shown to offer more accurate estimations of tracheobronchial dimensions, while ultrasound has been identified as a promising bedside tool. Recent meta-analytic evidence and technological advancements, including 3D reconstruction and AI-based modelling, may support a more personalised and safer approach. It is recommended that a pragmatic, image-guided strategy be employed to minimise airway trauma, improve lung isolation, and optimise patient outcomes.