Lung cancer remains the leading cause of cancer-related mortality worldwide, largely due to late-stage diagnosis. Early detection is critical for improving patient outcomes, yet current screening methods, such as low-dose computed tomography (CT), often lack the sensitivity and specificity required for early-stage detection. Here, we present a multimodal early screening platform that integrates a multiplexed laser-induced graphene (LIG) immunosensor with machine learning to enhance the accuracy of lung cancer diagnosis. Our platform enables the rapid, cost-effective, and simultaneous detection of four tumor markers─neuron-specific enolase (NSE), carcinoembryonic antigen (CEA), p53, and SOX2─with limits of detection (LOD) as low as 1.62 pg/mL. By combining proteomic data from the immunosensor with deep learning-based CT imaging features and clinical data, we developed a multimodal predictive model that achieves an area under the curve (AUC) of 0.936, significantly outperforming single-modality approaches. This platform offers a transformative solution for early lung cancer screening, particularly in resource-limited settings, and provides potential technical support for precision medicine in oncology.

Pneumonia detection in chest X-rays (CXR) increasingly relies on AI-driven diagnostic systems. However, their "black-box" nature often lacks transparency, underscoring the need for interpretability to improve patient outcomes. This study presents the first application of the Deformable Prototypical Part Network (D-ProtoPNet), an ante-hoc interpretable deep learning (DL) model, for pneumonia classification in pediatric patients' CXR images. Clinical insights were integrated through expert radiologist evaluation of the model's learned prototypes and activated image patches, ensuring that explanations aligned with medically meaningful features. The model was developed and tested on a retrospective dataset of 5,856 CXR images of pediatric patients, ages 1-5 years. The images were originally acquired at a tertiary academic medical center as part of routine clinical care and were publicly hosted on a Kaggle platform. This dataset comprised anterior-posterior images labeled normal, viral, and bacterial. It was divided into 80 % training and 20 % validation splits, and utilised in a supervised five-fold cross-validation. Performance metrics were compared with the original ProtoPNet, utilising ResNet50 as the base model. An experienced radiologist assessed the clinical relevance of the learned prototypes, patch activations, and model explanations. The D-ProtoPNet achieved an accuracy of 86 %, precision of 86 %, recall of 85 %, and AUC of 93 %, marking a 3 % improvement over the original ProtoPNet. While further optimisation is required before clinical use, the radiologist praised D-ProtoPNet's intuitive explanations, highlighting its interpretability and potential to aid clinical decision-making. Prototypical part learning offers a balance between classification performance and explanation quality, but requires improvements to match the accuracy of black-box models. This study underscores the importance of integrating domain expertise during model evaluation to ensure the interpretability of XAI models is grounded in clinically valid insights.

Introduction: Chest CT scans are increasingly used in dyspneic patients where acute heart failure (AHF) is a key differential diagnosis. Interpretation remains challenging and radiology reports are frequently delayed due to a radiologist shortage, although flagging such information for emergency physicians would have therapeutic implication. Artificial intelligence (AI) can be a complementary tool to enhance the diagnostic precision. We aim to develop an explainable AI model to detect radiological signs of AHF in chest CT with an accuracy comparable to thoracic radiologists. Methods: A single-center, retrospective study during 2016-2021 at Copenhagen University Hospital - Bispebjerg and Frederiksberg, Denmark. A Boosted Trees model was trained to predict AHF based on measurements of segmented cardiac and pulmonary structures from acute thoracic CT scans. Diagnostic labels for training and testing were extracted from radiology reports. Structures were segmented with TotalSegmentator. Shapley Additive explanations values were used to explain the impact of each measurement on the final prediction. Results: Of the 4,672 subjects, 49% were female. The final model incorporated twelve key features of AHF and achieved an area under the ROC of 0.87 on the independent test set. Expert radiologist review of model misclassifications found that 24 out of 64 (38%) false positives and 24 out of 61 (39%) false negatives were actually correct model predictions, with the errors originating from inaccuracies in the initial radiology reports. Conclusion: We developed an explainable AI model with strong discriminatory performance, comparable to thoracic radiologists. The AI model's stepwise, transparent predictions may support decision-making.

Recent work has revisited the infamous task Name that dataset and established that in non-medical datasets, there is an underlying bias and achieved high Accuracies on the dataset origin task. In this work, we revisit the same task applied to popular open-source chest X-ray datasets. Medical images are naturally more difficult to release for open-source due to their sensitive nature, which has led to certain open-source datasets being extremely popular for research purposes. By performing the same task, we wish to explore whether dataset bias also exists in these datasets. % We deliberately try to increase the difficulty of the task by dataset transformations. We apply simple transformations of the datasets to try to identify bias. Given the importance of AI applications in medical imaging, it's vital to establish whether modern methods are taking shortcuts or are focused on the relevant pathology. We implement a range of different network architectures on the datasets: NIH, CheXpert, MIMIC-CXR and PadChest. We hope this work will encourage more explainable research being performed in medical imaging and the creation of more open-source datasets in the medical domain. The corresponding code will be released upon acceptance.

<b>BACKGROUND</b>. Artificial intelligence (AI) tools for evaluating low-dose CT (LDCT) lung cancer screening examinations are used predominantly for assisting radiologists' interpretations. Alternate utilization scenarios (e.g., use of AI as a prescreener or backup) warrant consideration. <b>OBJECTIVE</b>. The purpose of this study was to evaluate the impact of different AI utilization scenarios on diagnostic outcomes and interpretation times for LDCT lung cancer screening. <b>METHODS</b>. This retrospective study included 366 individuals (358 men, 8 women; mean age, 64 years) who underwent LDCT from May 2017 to December 2017 as part of an earlier prospective lung cancer screening trial. Examinations were interpreted by one of five readers, who reviewed their assigned cases in two sessions (with and without a commercial AI computer-aided detection tool). These interpretations were used to reconstruct simulated AI utilization scenarios: as an assistant (i.e., radiologists interpret all examinations with AI assistance), as a prescreener (i.e., radiologists only interpret examinations with a positive AI result), or as backup (i.e., radiologists reinterpret examinations when AI suggests a missed finding). A group of thoracic radiologists determined the reference standard. Diagnostic outcomes and mean interpretation times were assessed. Decision-curve analysis was performed. <b>RESULTS</b>. Compared with interpretation without AI (recall rate, 22.1%; per-nodule sensitivity, 64.2%; per-examination specificity, 88.8%; mean interpretation time, 164 seconds), AI as an assistant showed higher recall rate (30.3%; <i>p</i> < .001), lower per-examination specificity (81.1%), and no significant change in per-nodule sensitivity (64.8%; <i>p</i> = .86) or mean interpretation time (161 seconds; <i>p</i> = .48); AI as a prescreener showed lower recall rate (20.8%; <i>p</i> = .02) and mean interpretation time (143 seconds; <i>p</i> = .001), higher per-examination specificity (90.3%; <i>p</i> = .04), and no significant difference in per-nodule sensitivity (62.9%; <i>p</i> = .16); and AI as a backup showed increased recall rate (33.6%; <i>p</i> < .001), per-examination sensitivity (66.4%; <i>p</i> < .001), and mean interpretation time (225 seconds; <i>p</i> = .001), with lower per-examination specificity (79.9%; <i>p</i> < .001). Among scenarios, only AI as a prescreener demonstrated higher net benefit than interpretation without AI; AI as an assistant had the least net benefit. <b>CONCLUSION</b>. Different AI implementation approaches yield varying outcomes. The findings support use of AI as a prescreener as the preferred scenario. <b>CLINICAL IMPACT</b>. An approach whereby radiologists only interpret LDCT examinations with a positive AI result can reduce radiologists' workload while preserving sensitivity.

Background: Accurate forecasting of lung tumor motion is essential for precise dose delivery in proton therapy. While current markerless methods mostly rely on deep learning, transformer-based architectures remain unexplored in this domain, despite their proven performance in trajectory forecasting. Purpose: This work introduces a markerless forecasting approach for lung tumor motion using Vision Transformers (ViT). Two training strategies are evaluated under clinically realistic constraints: a patient-specific (PS) approach that learns individualized motion patterns, and a multi-patient (MP) model designed for generalization. The comparison explicitly accounts for the limited number of images that can be generated between planning and treatment sessions. Methods: Digitally reconstructed radiographs (DRRs) derived from planning 4DCT scans of 31 patients were used to train the MP model; a 32nd patient was held out for evaluation. PS models were trained using only the target patient's planning data. Both models used 16 DRRs per input and predicted tumor motion over a 1-second horizon. Performance was assessed using Average Displacement Error (ADE) and Final Displacement Error (FDE), on both planning (T1) and treatment (T2) data. Results: On T1 data, PS models outperformed MP models across all training set sizes, especially with larger datasets (up to 25,000 DRRs, p < 0.05). However, MP models demonstrated stronger robustness to inter-fractional anatomical variability and achieved comparable performance on T2 data without retraining. Conclusions: This is the first study to apply ViT architectures to markerless tumor motion forecasting. While PS models achieve higher precision, MP models offer robust out-of-the-box performance, well-suited for time-constrained clinical settings.

To investigate the value of multi-model based on preoperative CT scans in predicting EGFR/TP53 co-mutation status. We retrospectively included 2171 patients with non-small cell lung cancer (NSCLC) with pre-treatment computed tomography (CT) scans and predicting epidermal growth factor receptor (EGFR) gene sequencing from West China Hospital between January 2013 and April 2024. The deep-learning model was built for predicting EGFR / tumor protein 53 (TP53) co-occurrence status. The model performance was evaluated by area under the curve (AUC) and Kaplan-Meier analysis. We further compared multi-dimension model with three one-dimension models separately, and we explored the value of combining clinical factors with machine-learning factors. Additionally, we investigated 546 patients with 56-panel next-generation sequencing and low-dose computed tomography (LDCT) to explore the biological mechanisms of radiomics. In our cohort of 2171 patients (1,153 males, 1,018 females; median age 60 years), single-dimensional models were developed using data from 1,055 eligible patients. The multi-dimensional model utilizing a Random Forest classifier achieved superior performance, yielding the highest AUC of 0.843 for predicting EGFR/TP53 co-mutations in the test set. The multi-dimensional model demonstrates promising potential for non-invasive prediction of EGFR and TP53 co-mutations, facilitating early and informed clinical decision-making in NSCLC patients at risk of treatment resistance.

The labor-intensive nature of medical data annotation presents a significant challenge for respiratory disease diagnosis, resulting in a scarcity of high-quality labeled datasets in resource-constrained settings. Moreover, patient privacy concerns complicate the direct sharing of local medical data across institutions, and existing centralized data-driven approaches, which rely on amounts of available data, often compromise data privacy. This study proposes a federated few-shot learning framework with privacy-preserving mechanisms to address the issues of limited labeled data and privacy protection in diagnosing respiratory diseases. In particular, a meta-stochastic gradient descent algorithm is proposed to mitigate the overfitting problem that arises from insufficient data when employing traditional gradient descent methods for neural network training. Furthermore, to ensure data privacy against gradient leakage, differential privacy noise from a standard Gaussian distribution is integrated into the gradients during the training of private models with local data, thereby preventing the reconstruction of medical images. Given the impracticality of centralizing respiratory disease data dispersed across various medical institutions, a weighted average algorithm is employed to aggregate local diagnostic models from different clients, enhancing the adaptability of a model across diverse scenarios. Experimental results show that the proposed method yields compelling results with the implementation of differential privacy, while effectively diagnosing respiratory diseases using data from different structures, categories, and distributions.

Recent works have revisited the infamous task ``Name That Dataset'', demonstrating that non-medical datasets contain underlying biases and that the dataset origin task can be solved with high accuracy. In this work, we revisit the same task applied to popular open-source chest X-ray datasets. Medical images are naturally more difficult to release for open-source due to their sensitive nature, which has led to certain open-source datasets being extremely popular for research purposes. By performing the same task, we wish to explore whether dataset bias also exists in these datasets. To extend our work, we apply simple transformations to the datasets, repeat the same task, and perform an analysis to identify and explain any detected biases. Given the importance of AI applications in medical imaging, it's vital to establish whether modern methods are taking shortcuts or are focused on the relevant pathology. We implement a range of different network architectures on the datasets: NIH, CheXpert, MIMIC-CXR and PadChest. We hope this work will encourage more explainable research being performed in medical imaging and the creation of more open-source datasets in the medical domain. Our code can be found here: https://github.com/eedack01/x_ray_ds_bias.

Recent works have revisited the infamous task ``Name That Dataset'', demonstrating that non-medical datasets contain underlying biases and that the dataset origin task can be solved with high accuracy. In this work, we revisit the same task applied to popular open-source chest X-ray datasets. Medical images are naturally more difficult to release for open-source due to their sensitive nature, which has led to certain open-source datasets being extremely popular for research purposes. By performing the same task, we wish to explore whether dataset bias also exists in these datasets. To extend our work, we apply simple transformations to the datasets, repeat the same task, and perform an analysis to identify and explain any detected biases. Given the importance of AI applications in medical imaging, it's vital to establish whether modern methods are taking shortcuts or are focused on the relevant pathology. We implement a range of different network architectures on the datasets: NIH, CheXpert, MIMIC-CXR and PadChest. We hope this work will encourage more explainable research being performed in medical imaging and the creation of more open-source datasets in the medical domain. Our code can be found here: https://github.com/eedack01/x_ray_ds_bias.