Artificial intelligence as a predictive tool for gastric cancer: Bridging innovation, clinical translation, and ethical considerations

Abstract

With gastric cancer ranking among the most prevalent and deadly malignancies worldwide, early detection and individualized prognosis remain essential for improving patient outcomes. This letter discusses recent advancements in artificial intelligence (AI)-driven predictive tools for gastric cancer, emphasizing a computed tomography-based radiomic model that achieved a predictive accuracy of area under the curve of 0.893 for treatment response in advanced cases undergoing neoadjuvant immunochemotherapy. AI offers promising avenues for predictive accuracy and personalized treatment planning in gastric oncology. Additionally, this letter highlights the comparison of these AI tools with traditional methodologies, demonstrating their potential to streamline clinical workflows and address existing gaps in risk stratification and early detection. Furthermore, this letter addresses the ethical considerations and the need for robust clinical-AI collaboration to achieve reliable, transparent, and unbiased outcomes. Strengthening cross-disciplinary efforts will be vital for the responsible and effective deployment of AI in this critical area of oncology.

Key Words: Gastric cancer; Artificial intelligence; Machine learning; Deep learning; Predictive models

Core Tip: Artificial intelligence (AI) is transforming gastric cancer management by enhancing early detection and individualized treatment planning through advanced predictive models. This letter highlights recent innovations in AI, such as a computed tomography-based radiomic model that predicts responses to neoadjuvant immunochemotherapy in advanced gastric cancer patients. By integrating AI-driven analysis with clinical factors, these tools offer substantial predictive accuracy, promising improved patient outcomes. However, to fully realize AI’s potential, ongoing collaboration is essential to address ethical, technical, and validation challenges, ensuring AI’s responsible integration into clinical oncology for effective, transparent, and patient-centered care.

TO THE EDITOR

Gastric cancer, a leading cause of cancer-related deaths, often presents at advanced stages due to limited early detection methods, resulting in poor prognosis and limited treatment options[1-3]. Current diagnostic methods, such as endoscopy and histopathological analysis, are often invasive, resource-intensive, and limited in their ability to provide comprehensive, personalized insights into disease progression. Recent developments in artificial intelligence (AI) have spurred research into predictive models that can provide earlier and more accurate risk assessments, potentially shifting the diagnostic and therapeutic paradigm for this disease. With capabilities that extend beyond traditional statistical methods, AI-driven models can analyze complex, multi-dimensional datasets, identifying predictive markers and patterns that support early diagnosis, prognostication, and personalized treatment planning[4-6] (Figure 1).

Figure 1 Workflow of the artificial intelligence predictive model for gastric cancer. AI: Artificial intelligence.

This bar chart compares the performance of different AI models in predicting gastric cancer. The models are evaluated based on three metrics: Area under the curve (AUC), sensitivity, and specificity. AUC measures the overall model performance, while sensitivity and specificity assess the model's ability to correctly identify true positive and true negative cases, respectively.

Despite its promise, integrating AI into clinical oncology faces multiple challenges, from technical and ethical hurdles to the need for effective model validation across diverse populations. AI models must be robust enough to handle the heterogeneity of gastric cancer presentations and treatment responses. Validation processes for these models commonly include performance metrics like sensitivity, specificity, and AUC. Additionally, external validation is performed using independent datasets from diverse cohorts to ensure the model’s reliability across varied clinical settings[5-7].

As evidenced by the above radiomic model[7], achieving high predictive accuracy relies on quality datasets and advanced validation methods, which are critical for bridging research and clinical use. The radiomic feature extraction process for computed tomography-based models typically involves segmenting the region of interest, extracting quantitative features (e.g., texture, shape, and intensity), and using feature selection algorithms to identify those most correlated with clinical outcomes. These steps ensure that extracted features meaningfully contribute to the model's predictive power (Figure 2).

Figure 2 Comparison of predictive accuracy metrics for artificial intelligence models in gastric cancer. AUC: Area under the curve.

This letter examines the state of AI in gastric cancer prediction and provides perspectives on both its clinical integration and ethical use. Addressing these topics is crucial for harnessing AI’s full potential while ensuring patient-centered care in the management of gastric cancer.

CURRENT ADVANCES AND CHALLENGES IN AI-DRIVEN PREDICTION MODELS FOR GASTRIC CANCER

Recent technological advancements

AI has shown notable success in oncology, particularly through ML and deep learning (DL) techniques that can predict patient outcomes by learning from large datasets[8-10]. For gastric cancer, these models have achieved promising accuracy in identifying high-risk individuals, predicting tumor aggressiveness, recurrence, and survival rates. For instance, convolutional neural networks and other DL models have demonstrated potential in analyzing endoscopic images, genomic data, and histopathology slides, aiding in both diagnosis and prognosis with levels of accuracy that rival traditional methods[11-13].

Moreover, predictive models trained on patient demographics, laboratory data, and genetic profiles have opened new possibilities for risk stratification and personalized treatment planning. Integrating data from multiple modalities (e.g., imaging, genetics, clinical) has enabled AI to provide a holistic assessment of a patient’s cancer status and potential progression, which could fundamentally change the timing and personalization of treatments[14-16].

Challenges to widespread adoption

However, translating these AI models from research into clinical practice is not without obstacles. One primary issue is data heterogeneity—variability in data quality, collection methods, and patient demographics across institutions can undermine model performance and generalizability. Additionally, the “black-box” nature of many AI algorithms, especially DL models, poses interpretability issues for clinicians, limiting trust in AI recommendations without a clear understanding of the decision-making process[17-19].

Another critical challenge is the need for high-quality, annotated datasets that are inclusive of diverse patient populations, considering that many AI models suffer from biases due to the limited representativeness of the training datasets. Models must be validated across multiple clinical centers to ensure that they perform accurately and consistently in real-world settings. Addressing these technical challenges will be necessary for achieving reliable, reproducible results in gastric cancer prediction[18-20].

CLINICAL INTEGRATION AND ETHICAL CONSIDERATIONS FOR AI IN GASTRIC CANCER PREDICTION

Integration into clinical practice

To make a meaningful impact on patient outcomes, AI systems must be effectively integrated into clinical workflows, not as replacements for clinicians but as collaborative tools that enhance diagnostic and therapeutic precision. Integration includes mechanisms for presenting predictions to healthcare providers in user-friendly formats, such as visual dashboards or risk scores, to facilitate decision-making. Clinicians can incorporate these insights into their treatment planning processes, ensuring alignment with patient-specific needs and clinical judgment.

AI-driven models can act as an “early warning system”, flagging high-risk patients for further investigation and supporting clinicians in making data-informed decisions. This integration requires tailored, user-friendly interfaces that allow clinicians to review AI-generated predictions and insights efficiently[6,9,15].

Collaboration between AI developers and clinical oncologists is also vital for refining models based on practical insights and ensuring that AI systems complement rather than complicate clinical workflows. Clear protocols for clinician-AI interactions, ongoing training, and feedback loops will be essential to align the capabilities of AI with the unique needs of gastric cancer management[8,12,19].

Ethical considerations and bias prevention

As AI models are introduced into healthcare, ethical considerations must take precedence, particularly regarding patient privacy, transparency, and bias prevention. AI algorithms must operate with transparency, allowing clinicians to understand the basis for predictions and identify potential errors or biases. Algorithms trained on biased data can lead to disparities in care, affecting demographic groups differently. This issue necessitates careful data curation and ongoing validation to safeguard against unintended biases[2,11,20].

Moreover, the privacy of patient data must be rigorously protected, as AI models often require access to sensitive information for training and validation. Clear guidelines are needed to ensure that patient data is anonymized and securely stored, especially when working with multi-center datasets. Cross-disciplinary efforts involving AI researchers, ethicists, and oncologists are crucial for creating robust frameworks that prioritize patient welfare and equitable healthcare access.

CONCLUSION

AI presents a transformative opportunity in the fight against gastric cancer, with its potential to enable earlier detection, improve risk stratification, and personalize treatment strategies. However, realizing this potential will require addressing the technical, clinical, and ethical challenges associated with its use. Building on the strengths of recent technological advances, the medical community must foster interdisciplinary collaborations that promote transparency, inclusivity, and trust in AI tools. By prioritizing ethical standards and strengthening clinician-AI integration, we can pave the way for AI to serve as a reliable, patient-centered adjunct in gastric oncology. This approach aligns with the overarching goal of enhancing clinical decision-making while ensuring that AI applications in healthcare are both equitable and responsibly managed. The adoption of AI in gastric cancer care is not only a technological shift but a step towards a more proactive and personalized approach in oncology, promising significant advancements in patient outcomes.