Chest X-rays (CXR) rank among the most conducted X-ray examinations. They often require repeat imaging due to inadequate quality, leading to increased radiation exposure and delays in patient care and diagnosis. This research assesses the efficacy of DenseNet121 and YOLOv8 neural networks in detecting suboptimal CXRs, which may minimise delays and enhance patient outcomes. The study included 3587 patients with a median age of 67 (0-102). It utilised an initial dataset comprising 10,000 CXRs randomly divided into a training subset (4000 optimal and 4000 suboptimal) and a validation subset (400 optimal and 400 suboptimal). The test subset (25 optimal and 25 suboptimal) was curated from the remaining images to provide adequate variation. Neural networks DenseNet121 and YOLOv8 were chosen due to their capabilities in image classification. DenseNet121 is a robust, well-tested model in the medical industry with high accuracy in object recognition. YOLOv8 is a cutting-edge commercial model targeted at all industries. Their performance was assessed via the area under the receiver operating curve (AUROC) and compared to radiologist classification, utilising the chi-squared test. DenseNet121 attained an AUROC of 0.97, while YOLOv8 recorded a score of 0.95, indicating a strong capability in differentiating between optimal and suboptimal CXRs. The alignment between radiologists and models exhibited variability, partly due to the lack of clinical indications. However, the performance was not statistically significant. Both AI models effectively classified chest X-ray quality, demonstrating the potential for providing radiographers with feedback to improve image quality. Notably, this was the first study to include both PA and lateral CXRs as well as paediatric cases and the first to evaluate YOLOv8 for this application.

The assessment of diaphragm function is crucial for effective clinical management and the prevention of complications associated with diaphragmatic dysfunction. However, current measurement methodologies rely on manual techniques that are susceptible to human error: How does the performance of an automatic diaphragm measurement system based on a segmentation neural network focusing on diaphragm thickness and excursion compare with existing methodologies? The proposed system integrates segmentation and parameter measurement, leveraging a newly established ultrasound diaphragm dataset. This dataset comprises B-mode ultrasound images and videos for diaphragm thickness assessment, as well as M-mode images and videos for movement measurement. We introduce a novel deep learning-based segmentation network, the Multi-ratio Dilated U-Net (MDRU-Net), to enable accurate diaphragm measurements. The system additionally incorporates a comprehensive implementation plan for automated measurement. Automatic measurement results are compared against manual assessments conducted by clinicians, revealing an average error of 8.12% in diaphragm thickening fraction measurements and a mere 4.3% average relative error in diaphragm excursion measurements. The results indicate overall minor discrepancies and enhanced potential for clinical detection of diaphragmatic conditions. Additionally, we design a user-friendly automatic measurement system for assessing diaphragm parameters and an accompanying method for measuring ultrasound-derived diaphragm parameters. In this paper, we constructed a diaphragm ultrasound dataset of thickness and excursion. Based on the U-Net architecture, we developed an automatic diaphragm segmentation algorithm and designed an automatic parameter measurement scheme. A comparative error analysis was conducted against manual measurements. Overall, the proposed diaphragm ultrasound segmentation algorithm demonstrated high segmentation performance and efficiency. The automatic measurement scheme based on this algorithm exhibited high accuracy, eliminating subjective influence and enhancing the automation of diaphragm ultrasound parameter assessment, thereby providing new possibilities for diaphragm evaluation.

Lung neuroendocrine tumours (NETs, also known as carcinoids) are rapidly rising in incidence worldwide but have unknown aetiology and limited therapeutic options beyond surgery. We conducted multi-omic analyses on over 300 lung NETs including whole-genome sequencing (WGS), transcriptome profiling, methylation arrays, spatial RNA sequencing, and spatial proteomics. The integration of multi-omic data provides definitive proof of the existence of four strikingly different molecular groups that vary in patient characteristics, genomic and transcriptomic profiles, microenvironment, and morphology, as much as distinct diseases. Among these, we identify a new molecular group, enriched for highly aggressive supra-carcinoids, that displays an immune-rich microenvironment linked to tumour--macrophage crosstalk, and we uncover an undifferentiated cell population within supra-carcinoids, explaining their molecular and behavioural link to high-grade lung neuroendocrine carcinomas. Deep learning models accurately identified the Ca A1, Ca A2, and Ca B groups based on morphology alone, outperforming current histological criteria. The characteristic tumour microenvironment of supra-carcinoids and the validation of a panel of immunohistochemistry markers for the other three molecular groups demonstrates that these groups can be accurately identified based solely on morphological features, facilitating their implementation in the clinical setting. Our proposed morpho-molecular classification highlights group-specific therapeutic opportunities, including DLL3, FGFR, TERT, and BRAF inhibitors. Overall, our findings unify previously proposed molecular classifications and refine the lung cancer map by revealing novel tumour types and potential treatments, with significant implications for prognosis and treatment decision-making.

Pulmonary Nodules (PNs) are a trend considered as the early manifestation of lung cancer. Among them, PNs that remain stable for more than two years or whose pathological results suggest not being lung cancer are considered benign PNs (BPNs), while PNs that conform to the growth pattern of tumors or whose pathological results indicate lung cancer are considered malignant PNs (MPNs). Currently, more than 90% of PNs detected by screening tests are benign, with a false positive rate of up to 96.4%. While a range of predictive models have been developed for the identification of MPNs, there are still some challenges in distinguishing between BPNs and MPNs. We included a total of 5197 patients for the case-control study according to the preset exclusion criteria and sample size. Among them, 4735 with BPNs and 2509 with MPNs were randomly divided into training, validation, and test sets according to a 7:1.5:1.5 ratio. Three widely applicable machine learning algorithms (Random Forests, Gradient Boosting Machine, and XGBoost) were used to screen the metrics, and then the corresponding predictive models were constructed using discriminative analysis, and the best performing model was selected as the target model. The model is internally validated with 10-fold cross validation and compared with PKUPH and Block models. We collated information from chest CT examinations performed from 2018 to 2021 in the physical examination population and found that the detection rate of PNs was 21.57% and showed an overall upward trend. The GMU_D model constructed by discriminative analysis based on machine learning screening features had an excellent discriminative performance (AUC = 0.866, 95% CI: 0.858-0.874), and higher accuracy than the PKUPH model (AUC = 0.559, 95% CI: 0.552-0.567) and the Block model (AUC = 0.823, 95% CI: 0.814-0.833). Moreover, the cross-validation results also exhibit excellent performance (AUC = 0.866, 95% CI: 0.858-0.874). The detection rate of PNs was 21.57% in the physical examination population undergoing chest CT. Meanwhile, based on real-world studies of PNs, a greater prediction tool was developed and validated that can be used to accurately distinguish between BPNs and MPNs with the excellent predictive performance and differentiation.

Annotation variability remains a substantial challenge in medical image segmentation, stemming from ambiguous imaging boundaries and diverse clinical expertise. Traditional deep learning methods producing single deterministic segmentation predictions often fail to capture these annotator biases. Although recent studies have explored multi-rater segmentation, existing methods typically focus on a single perspective -- either generating a probabilistic ``gold standard'' consensus or preserving expert-specific preferences -- thus struggling to provide a more omni view. In this study, we propose DiffOSeg, a two-stage diffusion-based framework, which aims to simultaneously achieve both consensus-driven (combining all experts' opinions) and preference-driven (reflecting experts' individual assessments) segmentation. Stage I establishes population consensus through a probabilistic consensus strategy, while Stage II captures expert-specific preference via adaptive prompts. Demonstrated on two public datasets (LIDC-IDRI and NPC-170), our model outperforms existing state-of-the-art methods across all evaluated metrics. Source code is available at https://github.com/string-ellipses/DiffOSeg .

BackgroundArtificial Intelligence (AI) has evolved through various trends, with different subfields gaining prominence over time. Currently, Conversational Artificial Intelligence (CAI)--particularly Generative AI--is at the forefront. CAI models are primarily focused on text-based tasks and are commonly deployed as chatbots. Recent advancements by OpenAI have enabled the integration of external, independently developed models, allowing chatbots to perform specialized, task-oriented functions beyond general language processing. ObjectiveThis study aims to develop a smart chatbot that integrates large language models (LLMs) from OpenAI with specialized domain-specific models, such as those used in medical image diagnostics. The system leverages transfer learning via Googles Teachable Machine to construct image-based classifiers and incorporates a diabetes detection model developed in TensorFlow.js. A key innovation is the chatbots ability to extract relevant parameters from user input, trigger the appropriate diagnostic model, interpret the output, and deliver responses in natural language. The overarching goal is to demonstrate the potential of combining LLMs with external models to build multimodal, task-oriented conversational agents. MethodsTwo image-based models were developed and integrated into the chatbot system. The first analyzes chest X-rays to detect viral and bacterial pneumonia. The second uses optical coherence tomography (OCT) images to identify ocular conditions such as drusen, choroidal neovascularization (CNV), and diabetic macular edema (DME). Both models were incorporated into the chatbot to enable image-based medical query handling. In addition, a text-based model was constructed to process physiological measurements for diabetes prediction using TensorFlow.js. The architecture is modular: new diagnostic models can be added without redesigning the chatbot, enabling straightforward functional expansion. ResultsThe findings demonstrate effective integration between the chatbot and the diagnostic models, with only minor deviations from expected behavior. Additionally, a stub function was implemented within the chatbot to schedule medical appointments based on the severity of a patients condition, and it was specifically tested with the OCT and X-ray models. ConclusionsThis study demonstrates the feasibility of developing advanced AI systems--including image-based diagnostic models and chatbot integration--by leveraging Artificial Intelligence as a Service (AIaaS). It also underscores the potential of AI to enhance user experiences in bioinformatics, paving the way for more intuitive and accessible interfaces in the field. Looking ahead, the modular nature of the chatbot allows for the integration of additional diagnostic models as the system evolves.

Interpretability can improve the safety, transparency and trust of AI models, which is especially important in healthcare applications where decisions often carry significant consequences. Mechanistic interpretability, particularly through the use of sparse autoencoders (SAEs), offers a promising approach for uncovering human-interpretable features within large transformer-based models. In this study, we apply Matryoshka-SAE to the radiology-specialised multimodal large language model, MAIRA-2, to interpret its internal representations. Using large-scale automated interpretability of the SAE features, we identify a range of clinically relevant concepts - including medical devices (e.g., line and tube placements, pacemaker presence), pathologies such as pleural effusion and cardiomegaly, longitudinal changes and textual features. We further examine the influence of these features on model behaviour through steering, demonstrating directional control over generations with mixed success. Our results reveal practical and methodological challenges, yet they offer initial insights into the internal concepts learned by MAIRA-2 - marking a step toward deeper mechanistic understanding and interpretability of a radiology-adapted multimodal large language model, and paving the way for improved model transparency. We release the trained SAEs and interpretations: https://huggingface.co/microsoft/maira-2-sae.

This study aims to explore the role of intra- and peri-tumoral radiomics features in tumor risk prediction, with a particular focus on the impact of peri-tumoral characteristics on the tumor microenvironment. A total of 133 patients, including 128 with thymomas and 5 with thymic carcinomas, were ultimately enrolled in this study. Based on the high- and low-risk classification, the cohort was divided into a training set (n = 93) and a testing set (n = 40) for subsequent analysis.Based on imaging data from these 133 patients, multiple radiomics prediction models integrating intra-tumoral and peritumoral features were developed. The data were sourced from patients treated at the Affiliated Hospital of Guangdong Medical University between 2015 and 2023, with all imaging obtained through preoperative CT scans. Radiomics feature extraction involved three primary categories: first-order features, shape features, and high-order features. Initially, the tumor's region of interest (ROI) was manually delineated using ITK-SNAP software. A custom Python algorithm was then used to automatically expand the peri-tumoral area, extracting features within 1 mm, 2 mm, and 3 mm zones surrounding the tumor. Additionally, considering the multimodal nature of the imaging data, image fusion techniques were incorporated to further enhance the model's ability to capture the tumor microenvironment. To build the radiomics models, selected features were first standardized using z-scores. Initial feature selection was performed using a t-test (p < 0.05), followed by Spearman correlation analysis to remove redundancy by retaining only one feature from each pair with a correlation coefficient ≥ 0.90. Subsequently, hierarchical clustering and the LASSO algorithm were applied to identify the most predictive features. These selected features were then used to train machine learning models, which were optimized on the training dataset and assessed for predictive performance. To further evaluate the effectiveness of these models, various statistical methods were applied, including DeLong's test, NRI, and IDI, to compare predictive differences among models. Decision curve analysis (DCA) was also conducted to assess the clinical applicability of the models. The results indicate that the IntraPeri1mm model performed the best, achieving an AUC of 0.837, with sensitivity and specificity at 0.846 and 0.84, respectively, significantly outperforming other models. SHAP value analysis identified several key features, such as peri_log_sigma_2_0_mm 3D_firstorder RootMeanSquared and intra_wavelet_LLL_firstorder Skewness, which made substantial contributions to the model's predictive accuracy. NRI and IDI analyses further confirmed the model's superior clinical applicability, and the DCA curve demonstrated robust performance across different thresholds. DeLong's test highlighted the statistical significance of the IntraPeri1mm model, underscoring its potential utility in radiomics research. Overall, this study provides a new perspective on tumor risk assessment, highlighting the importance of peri-tumoral features in the analysis of the tumor microenvironment. It aims to offer valuable insights for the development of personalized treatment plans. Not applicable.

Medical image synthesis presents unique challenges due to the inherent complexity and high-resolution details required in clinical contexts. Traditional generative architectures such as Generative Adversarial Networks (GANs) or Variational Auto Encoder (VAEs) have shown great promise for high-resolution image generation but struggle with preserving fine-grained details that are key for accurate diagnosis. To address this issue, we introduce Pixel Perfect MegaMed, the first vision-language foundation model to synthesize images at resolutions of 1024x1024. Our method deploys a multi-scale transformer architecture designed specifically for ultra-high resolution medical image generation, enabling the preservation of both global anatomical context and local image-level details. By leveraging vision-language alignment techniques tailored to medical terminology and imaging modalities, Pixel Perfect MegaMed bridges the gap between textual descriptions and visual representations at unprecedented resolution levels. We apply our model to the CheXpert dataset and demonstrate its ability to generate clinically faithful chest X-rays from text prompts. Beyond visual quality, these high-resolution synthetic images prove valuable for downstream tasks such as classification, showing measurable performance gains when used for data augmentation, particularly in low-data regimes. Our code is accessible through the project website - https://tehraninasab.github.io/pixelperfect-megamed.

Interpretability can improve the safety, transparency and trust of AI models, which is especially important in healthcare applications where decisions often carry significant consequences. Mechanistic interpretability, particularly through the use of sparse autoencoders (SAEs), offers a promising approach for uncovering human-interpretable features within large transformer-based models. In this study, we apply Matryoshka-SAE to the radiology-specialised multimodal large language model, MAIRA-2, to interpret its internal representations. Using large-scale automated interpretability of the SAE features, we identify a range of clinically relevant concepts - including medical devices (e.g., line and tube placements, pacemaker presence), pathologies such as pleural effusion and cardiomegaly, longitudinal changes and textual features. We further examine the influence of these features on model behaviour through steering, demonstrating directional control over generations with mixed success. Our results reveal practical and methodological challenges, yet they offer initial insights into the internal concepts learned by MAIRA-2 - marking a step toward deeper mechanistic understanding and interpretability of a radiology-adapted multimodal large language model, and paving the way for improved model transparency. We release the trained SAEs and interpretations: https://huggingface.co/microsoft/maira-2-sae.