Artificial Intelligence (AI) and deep learning models have revolutionized diagnosis, prognostication, and treatment planning by extracting complex patterns from medical images, enabling more accurate, personalized, and timely clinical decisions. Despite its promise, challenges such as image heterogeneity across different centers, variability in acquisition protocols and scanners, and sensitivity to artifacts hinder the reliability and clinical integration of deep learning models. Addressing these issues is critical for ensuring accurate and practical AI-powered neuroimaging applications. We reviewed and summarized the strategies for improving the robustness and generalizability of deep learning models for the segmentation and classification of neuroimages. This review follows a structured protocol, comprehensively searching Google Scholar, PubMed, and Scopus for studies on neuroimaging, task-specific applications, and model attributes. Peer-reviewed, English-language studies on brain imaging were included. The extracted data were analyzed to evaluate the implementation and effectiveness of these techniques. The study identifies key strategies to enhance deep learning in neuroimaging, including regularization, data augmentation, transfer learning, and uncertainty estimation. These approaches address major challenges such as data variability and domain shifts, improving model robustness and ensuring consistent performance across diverse clinical settings. The technical strategies summarized in this review can enhance the robustness and generalizability of deep learning models for segmentation and classification to improve their reliability for real-world clinical practice.

Accurate risk-stratification can facilitate precision therapy in oropharyngeal squamous cell carcinoma (OPSCC). We explored the potential added value of baseline positron emission tomography (PET)/computed tomography (CT) radiomic features for prognostication and risk stratification of OPSCC beyond the American Joint Committee on Cancer (AJCC) 8th edition staging scheme. Using institutional and publicly available datasets, we included OPSCC patients with known human papillomavirus (HPV) status, without baseline distant metastasis and treated with curative intent. We extracted 1037 PET and 1037 CT radiomic features quantifying lesion shape, imaging intensity, and texture patterns from primary tumors and metastatic cervical lymph nodes. Utilizing random forest algorithms, we devised novel machine-learning models for OPSCC progression-free survival (PFS) and overall survival (OS) using “radiomics” features, “AJCC” variables, and the “combined” set as input. We designed both single- (PET or CT) and combined-modality (PET/CT) models. Harrell’s C-index quantified survival model performance; risk stratification was evaluated in Kaplan–Meier analysis. A total of 311 patients were included. In HPV-associated OPSCC, the best “radiomics” model achieved an average C-index ± standard deviation of 0.62 ± 0.05 (p = 0.02) for PFS prediction, compared to 0.54 ± 0.06 (p = 0.32) utilizing “AJCC” variables. Radiomics-based risk-stratification of HPV-associated OPSCC was significant for PFS and OS. Similar trends were observed in HPV-negative OPSCC. In conclusion, radiomics imaging features extracted from pre-treatment PET/CT may provide complimentary information to the current AJCC staging scheme for survival prognostication and risk-stratification of HPV-associated OPSCC.

Glioma and brain metastasis can be difficult to distinguish on conventional magnetic resonance imaging (MRI) due to the similarity of imaging features in specific clinical circumstances. Multiple studies have investigated the use of machine learning (ML) models for non-invasive differentiation of glioma from brain metastasis. Many of the studies report promising classification results, however, to date, none have been implemented into clinical practice. After a screening of 12,470 studies, we included 29 eligible studies in our systematic review. From each study, we aggregated data on model design, development, and best classifiers, as well as quality of reporting according to the TRIPOD statement. In a subset of eligible studies, we conducted a meta-analysis of the reported AUC. It was found that data predominantly originated from single-center institutions (n = 25/29) and only two studies performed external validation. The median TRIPOD adherence was 0.48, indicating insufficient quality of reporting among surveyed studies. Our findings illustrate that despite promising classification results, reliable model assessment is limited by poor reporting of study design and lack of algorithm validation and generalizability. Therefore, adherence to quality guidelines and validation on outside datasets is critical for the clinical translation of ML for the differentiation of glioma and brain metastasis.

Technological innovation has enabled the development of machine learning (ML) tools that aim to improve the practice of radiologists. In the last decade, ML applications to neuro-oncology have expanded significantly, with the pre-operative prediction of glioma grade using medical imaging as a specific area of interest. We introduce the subject of ML models for glioma grade prediction by remarking upon the models reported in the literature as well as by describing their characteristic developmental workflow and widely used classifier algorithms. The challenges facing these models—including data sources, external validation, and glioma grade classification methods —are highlighted. We also discuss the quality of how these models are reported, explore the present and future of reporting guidelines and risk of bias tools, and provide suggestions for the reporting of prospective works. Finally, this review offers insights into next steps that the field of ML glioma grade prediction can take to facilitate clinical implementation.

Background: A major driver of individual variation in long-term outcomes following a large vessel occlusion (LVO) stroke is the degree of collateral arterial circulation. We aimed to develop and evaluate machine-learning models that quantify LVO collateral status using admission computed tomography angiography (CTA) radiomics. Methods: We extracted 1116 radiomic features from the anterior circulation territories from admission CTAs of 600 patients experiencing an acute LVO stroke. We trained and validated multiple machine-learning models for the prediction of collateral status based on consensus from two neuroradiologists as ground truth. Models were first trained to predict (1) good vs. intermediate or poor, or (2) good vs. intermediate or poor collateral status. Then, model predictions were combined to determine a three-tier collateral score (good, intermediate, or poor). We used the receiver operating characteristics area under the curve (AUC) to evaluate prediction accuracy. Results: We included 499 patients in training and 101 in an independent test cohort. The best-performing models achieved an averaged cross-validation AUC of 0.80 ± 0.05 for poor vs. intermediate/good collateral and 0.69 ± 0.05 for good vs. intermediate/poor, and AUC = 0.77 (0.67–0.87) and AUC = 0.78 (0.70–0.90) in the independent test cohort, respectively. The collateral scores predicted by the radiomics model were correlated with (rho = 0.45, p = 0.002) and were independent predictors of 3-month clinical outcome (p = 0.018) in the independent test cohort. Conclusions: Automated tools for the assessment of collateral status from admission CTA—such as the radiomics models described here—can generate clinically relevant and reproducible collateral scores to facilitate a timely treatment triage in patients experiencing an acute LVO stroke.

Intracerebral hemorrhage (ICH) and perihematomal edema (PHE) are key imaging markers of primary and secondary brain injury in hemorrhagic stroke. Accurate segmentation and quantification of ICH and PHE can help with prognostication and guide treatment planning. In this study, we combined Swin-Unet Transformers with nnU-NETv2 convolutional network for segmentation of ICH and PHE on non-contrast head CTs. We also applied test-time data augmentations to assess individual-level prediction uncertainty, ensuring high confidence in prediction. The model was trained on 1782 CT scans from a multicentric trial and tested in two independent datasets from Yale (n = 396) and University of Berlin Charité Hospital and University Medical Center Hamburg-Eppendorf (n = 943). Model performance was evaluated with the Dice coefficient and Volume Similarity (VS). Our dual Swin-nnUNET model achieved a median (95% confidence interval) Dice = 0.93 (0.90–0.95) and VS = 0.97 (0.95–0.98) for ICH, and Dice = 0.70 (0.64–0.75) and VS = 0.87 (0.80–0.93) for PHE segmentation in the Yale cohort. Dice = 0.86 (0.80–0.90) and VS = 0.91 (0.85–0.95) for ICH and Dice = 0.65 (0.56–0.70) and VS = 0.86 (0.77–0.93) for PHE segmentation in the Berlin/Hamburg-Eppendorf cohort. Prediction uncertainty was associated with lower segmentation accuracy, smaller ICH/PHE volumes, and infratentorial location. Our results highlight the benefits of a dual transformer-convolutional neural network architecture for ICH/PHE segmentation and test-time augmentation for uncertainty quantification.