Table 16.

Assessment of shared limitations and shared future perspectives for previous studies from sections 3.1–3.7.

Application	Shared limitation	Shared future perspective
Finding extraction from radiology report	1. Complex medical language processing: challenges in accurately processing and interpreting complex medical language structures	1. Advancements in medical language processing: innovations to improve accuracy in processing and interpreting complex medical language structures
Medical image/video captioning	2. Data limitations: challenges with dataset size, diversity, and quality affecting model performance and generalizability	2. Enhanced data resources: efforts to acquire larger, more diverse, and higher-quality datasets to enhance model training, performance, and generalizability
Diagnosis interpretability	3. Model training and stability: issues with the effectiveness and specificity of chosen models or methodologies, including labeling accuracy and classifier performance	3. Model training optimization: research on refining models and methodologies, focusing on effectiveness, specificity, labeling accuracy, and classifier performance
Medical report classification	4. Model specificity and language limitations: dependency on specific data formats, models, languages, or institutional settings, which may limit generalizability and impact versatility	4. Model generalization and adaptability: developing models and approaches that are less reliant on specific data formats, languages, or settings, enhancing generalizability and versatility
Medical report generation	5. Data diversity and availability: common issues include reliance on limited datasets, affecting model training and generalization	5. Promoting data diversity: initiatives to broaden datasets, ensuring they are more representative and comprehensive, improving model training and applicability
Multimodal learning
Medical visual question answering