Open research questions in COVID-19 diagnosis using AI

What the literature leaves open

• Current ILD segmentation methods have limitations such as cross-scenario generalization and fine-grained lesion differentiation. The limited receptive field of mainstream CNN-based methods cannot capture widely distributed reticulation and honeycomb lesions. The failure of Swin-UNet to extract diagonal strip-shaped ground-glass opacities results in jagged incomplete lesion contours.

  MCSA-Net: multi-scale convolutional slash attention network for interstitial lung disease segmentation · 2026 · DOI

• The lack of understanding of the effects of lung segmentation preprocessing on pneumonia classification in chest radiographs. The need to investigate the context-dependent effects of lung segmentation preprocessing across different acquisition settings and backbone architectures. The importance of considering the acquisition view and preprocessing strategy when developing pneumonia classification models.

  Context-Dependent Effects of Lung Segmentation Preprocessing on Radiology-Report-Derived Pneumonia Classification in Chest Radiographs: A Comparative Study Across Acquisition Settings and Backbone Architectures · 2026 · DOI

• Existing biomarkers for differentiation have limited accuracy. Laboratory testing methods suffer from limitations such as long turnaround times and low positive rates.

  FLIFuse: a lightweight parallel multi-branch network for pneumonia pathogen identification in chest x-rays · 2026 · DOI

• Conventional measurement methods are time-consuming and impractical. Portable oximeters have limitations, such as requiring physical contact and disposable probes.

  Estimation of Total Hemoglobin (SpHb) from Facial Videos Using 3D Convolutional Neural Network-Based Regression · 2025 · DOI

• The need to address variability in class representation and promote balanced learning.

  Lung Disease Classification Using Deep Learning and ROI-Based Chest X-Ray Images · 2025 · DOI

• However, generalization across institutions, imaging devices, patient populations, and acquisition protocols remains to be established through external validation.

  Progressive multi-scale learning for lightweight YOLO-based multi-disease detection in chest X-ray · 2026 · DOI

• Future research will focus on developing a clinically viable, human-in-the-loop diagnostic workflow, further optimizing performance through the integration of transformer-based language models to improve interpretability and decision-making.

  YOLOv8 framework for COVID-19 and pneumonia detection using synthetic image augmentation · 2025 · DOI

• The ResNet50 pre-trained architecture achieves identical 97.78% accuracy as CNN with F1-score 0.978, suggesting potential redundancy in the hybrid framework; ablation studies isolating the contribution of each component (CNN vs. ResNet50 vs. RNN vs. LSTM) and their interaction effects are needed to justify the four-model ensemble for lung disease recurrence prediction.

  Performance Analysis of a Hybrid Deep Learning Framework Integrating CNN, RNN, LSTM, and ResNet50 for Lung Disease Recurrence Prediction Using Chest X-Ray Images and Post-Recovery Clinical Data · 2026 · DOI

• Limited diagnostic resources and variability in manual Chest X-Ray interpretation. Need for accurate and efficient AI-assisted TB screening.

  Enhancing Tuberculosis Detection from Chest X-Ray Images Using Deep Learning: Evaluating Multi-Architecture Performance and Efficiency · 2026 · DOI

• The lack of access to conventional diagnostic measures in underdeveloped and developing nations. The need for a lightweight and deployable diagnostic-aid solution.

  A Convolutional Neural Network Ensemble Model for Pneumonia Detection using Chest X-Ray Images · 2026 · DOI

• The differentiation of COVID-19 cases from other respiratory conditions using CXR remains a complex task. Conventional diagnostic techniques have limitations, and there is a need for quick and accurate diagnosis methods.

  Leveraging deep residual learning and gradient boosting for COVID-19 detection in chest radiographs · 2026 · DOI

• Inconsistencies in pneumonia labeling and data quality issues remain challenges. The limited generalizability of models across different healthcare settings is a limitation. Data biases, model interpretability, and standardization are also challenges.

  Artificial Intelligence Applications in Pneumonia: Diagnosis and Outcome Prediction · 2026 · DOI

• The existing scoring systems for pneumonia outcome prediction do not account for complex interactions between clinical features or nonlinear risk patterns as observed in real-world clinical data. The limited generalizability of models across different healthcare settings is a gap.

  Artificial Intelligence Applications in Pneumonia: Diagnosis and Outcome Prediction · 2026 · DOI

• Predicting the progression of epidemics is crucial to minimize their impact. Existing models may not be effective in predicting the progression of epidemics. The study needs to evaluate the performance of RNN-based deep learning models.

  The RNN-Based Deep Learning Model Design to Predict ICU Occupancy Rate and Number of Intubated Patients for Effective Healthcare System Management · 2026 · DOI

• There is a need for predictive models that can accurately forecast the number of intensive care and intubated patients during a pandemic. Existing models may not be effective in predicting the progression of epidemics.

  The RNN-Based Deep Learning Model Design to Predict ICU Occupancy Rate and Number of Intubated Patients for Effective Healthcare System Management · 2026 · DOI

• Future work may investigate model compression techniques, efficient transformer variants, and lightweight architectures to further reduce computational cost and enable real-time deployment in clinical environments.

  Grad-CAM based deep learning analytics for image-level colon disease classification based on graph neural networks and vision transformers · 2026 · DOI

• Check marks indicate correct predictions. When the report generation outputs for both Scenario 1 (TR) and Scenario 2 (EN) are examined together, it becomes evident that the model can transform multi-slice visual representations into textual descriptions with high consistency in both languages.

  An Explainable Transformer-Based Framework for Lung Cancer Classification and Automated Radiology Report Generation from Multi-Slice CT Images · 2026 · DOI

• The model is trained and evaluated on a limited dataset, - The annotations were manually drawn by one of the authors, - The model's dependence on the training-only scribble prior may be…

  Seeing the subtle: lesion-aware multi-scale fusion for scribble-supervised segmentation · 2026 · DOI

• Investigating the application of LAMF to other medical image segmentation tasks, - Exploring the use of other types of supervision, such as box or polygon…

  Seeing the subtle: lesion-aware multi-scale fusion for scribble-supervised segmentation · 2026 · DOI

• Further breakthroughs are required for cross-scenario generalization and fine-grained lesion differentiation, - Research on FDA-approved chest radiology AI tools can further promote the clinical translation of ILD segmentation…

  MCSA-Net: multi-scale convolutional slash attention network for interstitial lung disease segmentation · 2026 · DOI

• Advancing technologies for robust and rapid diagnostic tools, - Bolstering preparedness against future epidemic threats, - Developing more sensitive and reliable detection…

  AI-Driven Comprehensive SERS-LFIA System: Improving Virus Automated Diagnostics Through SERS Image Recognition and Deep Learning · 2025 · DOI

• Further improvement of the model's performance, - Exploration of other applications of the 3D CNN-based regression…

  Estimation of Total Hemoglobin (SpHb) from Facial Videos Using 3D Convolutional Neural Network-Based Regression · 2025 · DOI

• The images contained in the database are very dependent on their origin, - Class weighting was not included in the final experimental pipeline, - The mean IoU remained low for all CAM-based explainability methods, - The study did not include near-duplicate…

  Lung Disease Classification Using Deep Learning and ROI-Based Chest X-Ray Images · 2025 · DOI

• The model's performance may be affected by the quality of the input images, - The model requires a large dataset for training, - The model's interpretability may be limited by the complexity of the explainability…

  An End-to-End Lightweight Multi-Scale CNN for the Classification of Lung and Colon Cancer with XAI Integration · 2024 · DOI

• Incorporating explainable AI methods. Statistically robust validation strategies. Evaluation on independent and diverse clinical datasets.

  Enhancing Tuberculosis Detection from Chest X-Ray Images Using Deep Learning: Evaluating Multi-Architecture Performance and Efficiency · 2026 · DOI

Most-cited papers in COVID-19 diagnosis using AI

• Automated detection of COVID-19 cases using deep neural networks with X-ray images · Computers in Biology and Medicine · 2020 · 2,183 citations
• COVID-Net: a tailored deep convolutional neural network design for detection of COVID-19 cases from chest X-ray images · Scientific Reports · 2020 · 1,881 citations
• Covid-19: automatic detection from X-ray images utilizing transfer learning with convolutional neural networks · Physical and Engineering Sciences in Medicine · 2020 · 1,827 citations
• Can AI Help in Screening Viral and COVID-19 Pneumonia? · IEEE Access · 2020 · 1,618 citations
• Using Artificial Intelligence to Detect COVID-19 and Community-acquired Pneumonia Based on Pulmonary CT: Evaluation of the Diagnostic Accuracy · Radiology · 2020 · 1,538 citations
• Deep learning-enabled medical computer vision · npj Digital Medicine · 2021 · 1,269 citations
• Automatic detection of coronavirus disease (COVID-19) using X-ray images and deep convolutional neural networks · Pattern Analysis and Applications · 2021 · 1,235 citations
• A Review of Deep Learning in Medical Imaging: Imaging Traits, Technology Trends, Case Studies With Progress Highlights, and Future Promises · Proceedings of the IEEE · 2021 · 1,142 citations
• Artificial intelligence: A powerful paradigm for scientific research · The Innovation · 2021 · 1,130 citations
• Inf-Net: Automatic COVID-19 Lung Infection Segmentation From CT Images · IEEE Transactions on Medical Imaging · 2020 · 1,081 citations

Most recent work

• Convolutional Fourier Analysis Network (CONV-FAN-POX): A novel Time–Frequency approach for medical image analysis · Biomedical Signal Processing and Control · 2026
• Multimodal artificial intelligence for disease diagnosis: Advances, applications, and challenges · Pattern Recognition · 2026
• A Hybrid of ResNext101_32x8d and Swin Transformer Networks with XAI for Alzheimer’s Disease Detection · Computers · 2026
• DB-HDFFN: Dual Branch Hierarchical Dynamic Feature Fusion Network for medical image classification · Biomedical Signal Processing and Control · 2026
• Foundation models in radiology: a primer for pediatric radiologists · Pediatric Radiology · 2026
• Lite-XNet: An efficient deep learning model for COVID-19 lung severity assessment on embedded systems · Computer Vision and Image Understanding · 2026
• RoentMod: a synthetic chest X-ray modification model to identify and correct image interpretation model shortcuts · npj Digital Medicine · 2026
• MedFusionT5: Cross-Modal Attention Boosts Semantic Quality and Reduces Hallucinations in Dental AI · International Dental Journal · 2026
• Artificial Intelligence-Assisted Pathogen Detection: Algorithms, Biosensing Platforms, and Applications · Biosensors · 2026
• Large-Scale Foundation Models for Radiological Image Analysis: Clinical Applications, Technical Challenges, and Future Directions · Journal of Imaging Informatics in Medicine · 2026

Related topics in Medicine