Table 2.

Comparison and summarization of the strengths and weaknesses of various AI-based multimodal tools across the different studies mentioned.

Field	Study	Strengths	Weaknesses
Radiology	Zhu et al., 2024 (Int J Surg) [28]	Strong in interpreting radiological images, handles large image–text datasets.	Limited medical domain expertise; potential for misinterpretation of complex cases.
Mental health	Elyoseph et al., 2024 (JMIR Ment Health) [29]	Effective at recognizing emotions from visual and textual data, helpful in mental health applications.	Potential for biases in emotion recognition and difficulty with more nuanced emotional contexts.
Neuroimaging	Biessmann et al., 2011 (IEEE Rev Biomed Eng) [30]	Robust integration of different neuroimaging modalities; enhances understanding of brain function.	Requires significant computational resources and expertise; challenges with data standardization.
Radiology	Brin et al., 2024 (Eur Radiol) [31]	Accurate interpretation of radiological images, effective across diverse clinical cases.	Performance varies with image complexity; potential for hallucination of results in complex imaging cases.
Ophthalmology	Sorin et al., 2024 (MedRxiv) [36]	Analysis of external ocular images with or without clinical context.	Performance was inferior to non-ophthalmologist physicians, and was only evaluated based on external images versus OCT or fundoscopy.
Pulmonary	Cahan et al., 2023 (Sci Rep) [32]	Improves mortality prediction by fusing clinical and imaging data, supports personalized treatment planning.	Model complexity and interpretability challenges; requires large, high-quality datasets.
Biomedicine	Acosta et al., 2022 (Nat Med) [33]	Integrates diverse medical data (imaging, clinical) for diagnostics, facilitates personalized care.	High dependency on comprehensive, high-quality datasets; concerns about generalizability in clinical practice.
Virtual reality in healthcare	Kouijzer et al., 2023 (Implement Sci Commun) [34]	Enhances patient engagement, effective for rehabilitation and training.	Implementation challenges, especially in integrating VR with existing healthcare systems.
Neurology	Ramanarayanan, 2024 (J Speech Lang Hear Res) [35]	Effective for remote monitoring and assessments, supports telehealth initiatives.	Privacy concerns and limited accuracy for certain complex conditions.
Immunology	Indolfi et al., 2024 (Front Med) [37]	Supports continuity of care from childhood to adulthood, personalized treatment recommendations.	Limited data on long-term outcomes; potential for biases in decision making.
Operations and patient management	Zhu et al., 2022 (Front Surg) [38]	Improves operational efficiency, integrates multimodal ID technology for patient and material management.	Complex integration with existing hospital systems; steep learning curve for users.
Radiology	Mohsen et al., 2022 (Sci Rep) [40]	Effective in combining EHR and imaging data for enhanced decision making, predictive analytics capabilities.	Requires high-quality, standardized datasets; concerns about patient privacy and data security.
Oncology	Lipkova et al., 2022 (Cancer Cell) [41]	Enables comprehensive analysis of multimodal cancer data (genomics, imaging), supports precision oncology.	Computationally expensive; challenges in scaling for real-time clinical applications.
Oncology/genomics	Shao et al., 2023 (Semin Cancer Biol) [42]	Enhances precision in predicting gene mutations through multimodal integration of genomics and clinical data.	Model complexity may hinder clinical interpretability; requires extensive training datasets.
Mental health	Alhuwaydi, 2024 (Risk Manag Healthc Policy) [43]	Improves accessibility to mental healthcare, supports early diagnosis and intervention through multimodal data.	Ethical concerns, potential biases in AI-driven mental health interventions, and lack of human empathy.
Mental health	Ettman & Galea, 2023 (JMIR Ment Health) [48]	Enhances population mental health monitoring, addresses large-scale public health issues with AI interventions.	Risks of widening socioeconomic disparities through AI implementation; data privacy concerns.