The Role of Artificial Intelligence Algorithms in Challenging Diagnostic Cases - Between Potential and Real Clinical Support.
Extranodal natural killer/T-cell lymphoma, nasal type (ENKTL), is a rare EBV-associated malignancy characterized by destructive tumors in the nasal cavity, often causing delayed diagnosis due to tissue necrosis. Artificial intelligence (AI) has potential to assist in complex diagnostic challenges.
This study aimed to evaluate the diagnostic performance and clinical relevance of publicly accessible AI-driven platforms in identifying ENKTL based on a real, challenging clinical case.
Clinical data from a 26-year-old male patient with protracted and complicated symptoms initially suggesting infectious conditions were used. Eight AI symptom assessment platforms were tested for diagnostic outputs using identical anonymized case inputs at early and advanced disease stages. Diagnostic suggestions, alignment with the final histopathological diagnosis, and quality of recommendations were analyzed.
Six of eight platforms provided valid differential diagnoses consistent with clinical data. Early-stage data mainly generated infectious diagnoses like peritonsillar abscess. Advanced-stage data shifted diagnoses toward invasive fungal, autoimmune, and malignant diseases, with platforms such as Doctronic and Symptomate explicitly including lymphoma. Diagnostic pathways varied, with recommendations for comprehensive laboratory tests, imaging, biopsies, and immunological assays. AI systems demonstrated oncologic vigilance when EBV viral data were included, emphasizing clinician oversight for accurate diagnosis.
AI-based diagnostic platforms show promise in supporting preliminary assessment and differential diagnosis in rare, complex cases such as ENKTL. While AI can guide diagnostic reasoning and propose investigations, it does not replace clinical judgment. Further development and validation of AI tools are needed to enhance their reliability and integration into medical practice.
This study aimed to evaluate the diagnostic performance and clinical relevance of publicly accessible AI-driven platforms in identifying ENKTL based on a real, challenging clinical case.
Clinical data from a 26-year-old male patient with protracted and complicated symptoms initially suggesting infectious conditions were used. Eight AI symptom assessment platforms were tested for diagnostic outputs using identical anonymized case inputs at early and advanced disease stages. Diagnostic suggestions, alignment with the final histopathological diagnosis, and quality of recommendations were analyzed.
Six of eight platforms provided valid differential diagnoses consistent with clinical data. Early-stage data mainly generated infectious diagnoses like peritonsillar abscess. Advanced-stage data shifted diagnoses toward invasive fungal, autoimmune, and malignant diseases, with platforms such as Doctronic and Symptomate explicitly including lymphoma. Diagnostic pathways varied, with recommendations for comprehensive laboratory tests, imaging, biopsies, and immunological assays. AI systems demonstrated oncologic vigilance when EBV viral data were included, emphasizing clinician oversight for accurate diagnosis.
AI-based diagnostic platforms show promise in supporting preliminary assessment and differential diagnosis in rare, complex cases such as ENKTL. While AI can guide diagnostic reasoning and propose investigations, it does not replace clinical judgment. Further development and validation of AI tools are needed to enhance their reliability and integration into medical practice.