Current medical image analysis systems are typically task-specific, requiring separate models for classification and segmentation, and lack the flexibility to support user-defined workflows. To address these challenges, we introduce MedPrompt, a unified framework that combines a few-shot prompted Large Language Model (Llama-4-17B) for high-level task planning with a modular Convolutional Neural Network (DeepFusionLab) for low-level image processing. The LLM interprets user instructions and generates structured output to dynamically route task-specific pretrained weights. This weight routing approach avoids retraining the entire framework when adding new tasks-only task-specific weights are required, enhancing scalability and deployment. We evaluated MedPrompt across 19 public datasets, covering 12 tasks spanning 5 imaging modalities. The system achieves a 97% end-to-end correctness in interpreting and executing prompt-driven instructions, with an average inference latency of 2.5 seconds, making it suitable for near real-time applications. DeepFusionLab achieves competitive segmentation accuracy (e.g., Dice 0.9856 on lungs) and strong classification performance (F1 0.9744 on tuberculosis). Overall, MedPrompt enables scalable, prompt-driven medical imaging by combining the interpretability of LLMs with the efficiency of modular CNNs.

Spiculation, characterized by irregular, spike-like projections from nodule margins, serves as a crucial radiological biomarker for malignancy assessment and early cancer detection. These distinctive stellate patterns strongly correlate with tumor invasiveness and are vital for accurate diagnosis and treatment planning. Traditional computer-aided diagnosis (CAD) systems are limited in their capability to capture and use these patterns given their subtlety, difficulty in quantifying them, and small datasets available to learn these patterns. To address these challenges, we propose a novel framework leveraging variational autoencoders (VAE) to discover, extract, and vary disentangled latent representations of lung nodule images. By gradually varying the latent representations of non-spiculated nodule images, we generate augmented datasets containing spiculated nodule variations that, we hypothesize, can improve the diagnostic classification of lung nodules. Using the National Institutes of Health/National Cancer Institute Lung Image Database Consortium (LIDC) dataset, our results show that incorporating these spiculated image variations into the classification pipeline significantly improves spiculation detection performance up to 7.53%. Notably, this enhancement in spiculation detection is achieved while preserving the classification performance of non-spiculated cases. This approach effectively addresses class imbalance and enhances overall classification outcomes. The gradual attenuation of spiculation characteristics demonstrates our model's ability to both capture and generate clinically relevant semantic features in an algorithmic manner. These findings suggest that the integration of semantic-based latent representations into CAD models not only enhances diagnostic accuracy but also provides insights into the underlying morphological progression of spiculated nodules, enabling more informed and clinically meaningful AI-driven support systems.

To develop a high-performance machine learning model for predicting and interpreting features of pulmonary diseases. This retrospective study analyzed clinical and imaging data from patients with non-small cell lung cancer (NSCLC), granulomatous inflammation, and benign tumors, collected across multiple centers from January 2015 to October 2023. Data from two hospitals in Anhui Province were split into a development set (n = 1696) and a test set (n = 424) in an 8:2 ratio, with an external validation set (n = 909) from Zhejiang Province. Features with p < 0.05 from univariate analyses were selected using the Boruta algorithm for input into Random Forest (RF) and XGBoost models. Model efficacy was assessed using receiver operating characteristic (ROC) analysis. A total of 3030 patients were included: 2269 with NSCLC, 529 with granulomatous inflammation, and 232 with benign tumors. The Obuchowski indices for RF and XGBoost in the test set were 0.7193 (95% CI: 0.6567-0.7812) and 0.8282 (95% CI: 0.7883-0.8650), respectively. In the external validation set, indices were 0.7932 (95% CI: 0.7572-0.8250) for RF and 0.8074 (95% CI: 0.7740-0.8387) for XGBoost. XGBoost achieved better accuracy in both the test (0.81) and external validation (0.79) sets. Calibration Curve and Decision Curve Analysis (DCA) showed XGBoost offered higher net clinical benefit. The XGBoost model outperforms RF in the three-class classification of lung diseases. XGBoost surpasses Random Forest in accurately classifying NSCLC, granulomatous inflammation, and benign tumors, offering superior clinical utility via multicenter data. Lung cancer classification model has broad clinical applicability. XGBoost outperforms random forests using CT imaging data. XGBoost model can be deployed on a website for clinicians.

To investigate automated breast ultrasound (ABUS) features affecting the use of Multiview convolutional neural network (CNN) for breast lesions according to radiologists' experience. A total of 656 breast lesions (152 malignant and 504 benign lesions) were included and reviewed by six radiologists for background echotexture, glandular tissue component (GTC), and lesion type and size without as well as with Multiview CNN. The sensitivity, specificity, and the area under the receiver operating curve (AUC) for ABUS features were compared between two sessions according to radiologists' experience. Radiology residents showed significant AUC improvement with the Multiview CNN for mass (0.81 to 0.91, P=0.003) and non-mass lesions (0.56 to 0.90, P=0.007), all background echotextures (homogeneous-fat: 0.84 to 0.94, P=0.04; homogeneous-fibroglandular: 0.85 to 0.93, P=0.01; heterogeneous: 0.68 to 0.88, P=0.002), all GTC levels (minimal: 0.86 to 0.93, P=0.001; mild: 0.82 to 0.94, P=0.003; moderate: 0.75 to 0.88, P=0.01; marked: 0.68 to 0.89, P<0.001), and lesions ≤10mm (≤5 mm: 0.69 to 0.86, P<0.001; 6-10 mm: 0.83 to 0.92, P<0.001). Breast specialists showed significant AUC improvement with the Multiview CNN in heterogeneous echotexture (0.90 to 0.95, P=0.03), marked GTC (0.88 to 0.95, P<0.001), and lesions ≤10mm (≤5 mm: 0.89 to 0.93, P=0.02; 6-10 mm: 0.95 to 0.98, P=0.01). With the Multiview CNN, the performance of ABUS in radiology residents was improved regardless of lesion type, background echotexture, or GTC. For breast lesions smaller than 10 mm, both radiology residents and breast specialists showed better performance of ABUS.

Post-surgical prediction of recurrence or metastasis for primary gastrointestinal stromal tumors (GISTs) remains challenging. We aim to develop individualized clinical follow-up strategies for primary GIST patients, such as shortening follow-up time or extending drug administration based on the clinical deep learning radiomics model (CDLRM). The clinical information on primary GISTs was collected from two independent centers. Postoperative recurrence or metastasis in GIST patients was defined as the endpoint of the study. A total of nine machine learning models were established based on the selected features. The performance of the models was assessed by calculating the area under the curve (AUC). The CDLRM with the best predictive performance was constructed. Decision curve analysis (DCA) and calibration curves were analyzed separately. Ultimately, our model was applied to the high-potential malignant group vs the low-malignant-potential group. The optimal clinical application scenarios of the model were further explored by comparing the DCA performance of the two subgroups. A total of 526 patients, 260 men and 266 women, with a mean age of 62 years, were enrolled in the study. CDLRM performed excellently with AUC values of 0.999, 0.963, and 0.995 for the training, external validation, and aggregated sets, respectively. The calibration curve indicated that CDLRM was in good agreement between predicted and observed probabilities in the validation cohort. The results of DCA's performance in different subgroups show that it was more clinically valuable in populations with high malignant potential. CDLRM could help the development of personalized treatment and improved follow-up of patients with a high probability of recurrence or metastasis in the future. This model utilizes imaging features extracted from CT scans (including radiomic features and deep features) and clinical data to accurately predict postoperative recurrence and metastasis in patients with primary GISTs, which has a certain auxiliary role in clinical decision-making. We developed and validated a model to predict recurrence or metastasis in patients taking oral imatinib after GIST. We demonstrate that CT image features were associated with recurrence or metastases. The model had good predictive performance and clinical benefit.

Diagnosis of tumefactive demyelination can be challenging. The diagnosis of indeterminate brain lesions on MRI often requires tissue confirmation via brain biopsy. Noninvasive methods for accurate diagnosis of tumor and nontumor etiologies allows for tailored therapy, optimal tumor control, and a reduced risk of iatrogenic morbidity and mortality. Tumefactive demyelination has imaging features that mimic <i>isocitrate dehydrogenase</i> wild-type glioblastoma (<i>IDH</i>wt GBM). We hypothesized that deep learning applied to postcontrast T1-weighted (T1C) and T2-weighted (T2) MRI can discriminate tumefactive demyelination from <i>IDH</i>wt GBM. Patients with tumefactive demyelination (<i>n</i> = 144) and <i>IDH</i>wt GBM (<i>n</i> = 455) were identified by clinical registries. A 3D DenseNet121 architecture was used to develop models to differentiate tumefactive demyelination and <i>IDH</i>wt GBM by using both T1C and T2 MRI, as well as only T1C and only T2 images. A 3-stage design was used: 1) model development and internal validation via 5-fold cross validation by using a sex-, age-, and MRI technology-matched set of tumefactive demyelination and <i>IDH</i>wt GBM, 2) validation of model specificity on independent <i>IDH</i>wt GBM, and 3) prospective validation on tumefactive demyelination and <i>IDH</i>wt GBM. Stratified area under the receiver operating curves (AUROCs) were used to evaluate model performance stratified by sex, age at diagnosis, MRI scanner strength, and MRI acquisition. The deep learning model developed by using both T1C and T2 images had a prospective validation AUROC of 88% (95% CI: 0.82-0.95). In the prospective validation stage, a model score threshold of 0.28 resulted in 91% sensitivity of correctly classifying tumefactive demyelination and 80% specificity (correctly classifying <i>IDH</i>wt GBM). Stratified AUROCs demonstrated that model performance may be improved if thresholds were chosen stratified by age and MRI acquisition. MRI can provide the basis for applying deep learning models to aid in the differential diagnosis of brain lesions. Further validation is needed to evaluate how well the model generalizes across institutions, patient populations, and technology, and to evaluate optimal thresholds for classification. Next steps also should incorporate additional tumor etiologies such as CNS lymphoma and brain metastases.

Pneumonia remains a significant cause of hospital mortality, prompting the need for precise mortality prediction methods. This study conducted a systematic review identifying predictors of mortality using Machine Learning (ML) and applied these methods to hospitalized pneumonia patients at the University Clinical Centre Zvezdara. The systematic review identified 16 studies (313,572 patients), revealing common mortality predictors including age, oxygen levels, and albumin. A Random Forest (RF) model was developed using local data (n=343), achieving an accuracy of 99%, and AUC of 0.99. Key predictors identified were chest X-ray worsening, ventilator use, age, and oxygen support. ML demonstrated high potential for accurately predicting pneumonia mortality, surpassing traditional severity scores, and highlighting its practical clinical utility.

This paper presents the results of a scoping review that examines potentials of Artificial Intelligence (AI) in early diagnosis of Cognitive Decline (CD), which is regarded as a key issue in elderly health. The review encompasses peer-reviewed publications from 2020 to 2025, including scientific journals and conference proceedings. Over 70% of the studies rely on using magnetic resonance imaging (MRI) as the input to the AI models, with a high diagnostic accuracy of 98%. Integration of the relevant clinical data and electroencephalograms (EEG) with deep learning methods enhances diagnostic accuracy in the clinical settings. Recent studies have also explored the use of natural language processing models for detecting CD at its early stages, with an accuracy of 75%, exhibiting a high potential to be used in the appropriate pre-clinical environments.

To develop and validate a computerized tomography (CT)‑based deep transfer learning radiomics model combined with explainable machine learning for preoperative risk prediction of thymoma. This retrospective study included 173 pathologically confirmed thymoma patients from our institution in the training group and 93 patients from two external centers in the external validation group. Tumors were classified according to the World Health Organization simplified criteria as low‑risk types (A, AB, and B1) or high‑risk types (B2 and B3). Radiomics features and deep transfer learning features were extracted from venous‑phase contrast‑enhanced CT images by using a modified Inception V3 network. Principal component analysis and least absolute shrinkage and selection operator regression identified 20 key predictors. Six classifiers-decision tree, gradient boosting machine, k‑nearest neighbors, naïve Bayes, random forest (RF), and support vector machine-were trained on five feature sets: CT imaging model, radiomics feature model, deep transfer learning feature model, combined feature model, and combined model. Interpretability was assessed with SHapley Additive exPlanations (SHAP), and an interactive web application was developed for real‑time individualized risk prediction and visualization. In the external validation group, the RF classifier achieved the highest area under the receiver operating characteristic curve (AUC) value of 0.956. In the training group, the AUC values for the CT imaging model, radiomics feature model, deep transfer learning feature model, combined feature model, and combined model were 0.684, 0.831, 0.815, 0.893, and 0.910, respectively. The corresponding AUC values in the external validation group were 0.604, 0.865, 0.880, 0.934, and 0.956, respectively. SHAP visualizations revealed the relative contribution of each feature, while the web application provided real‑time individual prediction probabilities with interpretative outputs. We developed a CT‑based deep transfer learning radiomics model combined with explainable machine learning and an interactive web application; this model achieved high accuracy and transparency for preoperative thymoma risk stratification, facilitating personalized clinical decision‑making.

The Thrombolysis in Cerebral Infarction (TICI) scale is widely used to assess angiographic outcomes of mechanical thrombectomy despite significant variability. Our objective was to create and optimize an artificial intelligence (AI)-based classification model for digital subtraction angiography (DSA) TICI scoring. Using a monocentric DSA dataset of thrombectomies, and a platform for medical image analysis, independent readers labeled each series according to TICI score and marked each thrombus. A convolutional neural network (CNN) classification model was created to classify TICI scores, into 2 groups (TICI 0,1 or 2a versus TICI 2b, 2c or 3) and 3 groups (TICI 0,1 or 2a versus TICI 2b versus TICI 2c or 3). The algorithm was first tested alone, and then thrombi positions were introduced to the algorithm by manual placement firstly, then after using a thrombus detection module. A total of 422 patients were enrolled in the study. 2492 thrombi were annotated on the TICI-labeled series. The model trained on a total of 1609 DSA series. The classification model into two classes had a specificity of 0.97 ±0.01 and a sensibility of 0.86 ±0.01. The 3-class models showed insufficient performance, even when combined with the true thrombi positions, with, respectively, F1 scores for TICI 2b classification of 0.50 and 0.55 ±0.07. The automatic thrombus detection module did not enhance the performance of the 3-class model, with a F1 score for the TICI 2b class measured at 0.50 ±0.07. The AI model provided a reproducible 2-class (TICI 0,1 or 2a versus 2b, 2c or 3) classification according to TICI scale. Its performance in distinguishing three classes (TICI 0,1 or 2a versus 2b versus 2c or 3) remains insufficient for clinical practice. Automatic thrombus detection did not improve the model's performance.