What question did this study set out to answer?

To assess the effectiveness of AI-driven models for predicting treatment responses in esophageal adenocarcinoma.

April 4, 2026Open Access

32P Whole-slide histopathology embeddings enable non-genomic HER2 biomarker prediction in esophageal adenocarcinoma

Key Points

To assess the effectiveness of AI-driven models for predicting treatment responses in esophageal adenocarcinoma.
Conducted a systematic review and meta-analysis of AI applications in cancer treatment predictions.
Evaluated predictive models for chemotherapy, immunotherapy, and targeted therapies.
Included 32 studies with over 12,000 patients published from 2015 to 2023.
AI models showed 75% predictive accuracy for treatment responses across cancer types.
Identified patients likely to benefit from immunotherapy with 65% sensitivity (p<0.001).
Achieved 72% specificity for predicting chemotherapy resistance (p<0.01).
In 2,500 breast cancer patients, a deep learning model achieved 80% accuracy for predicting neoadjuvant chemotherapy responses.

Abstract

Background: Artificial Intelligence (AI) and digital therapeutics (DTx) have emerged as transformative technologies in oncology, particularly in predicting and optimizing treatment responses.AI, through machine learning (ML) and deep learning (DL), has revolutionized early detection, precision medicine, and therapeutic decision-making.This study aimed to assess the efficacy of AI-driven predictive models in predicting cancer treatment responses, with a focus on breast cancer.Methods: A systematic review and meta-analysis were conducted, evaluating AI applications in predicting treatment responses for chemotherapy, immunotherapy, and targeted therapies.Studies were selected based on their use of AI models, including ML and DL algorithms, and clinical validation in prospective cohorts or randomized controlled trials (RCTs).A total of 32 studies published between 2015 and 2023, involving over 12,000 patients, were included.Results: AI-driven models demonstrated 75% predictive accuracy for treatment response across multiple cancer types, with 65% sensitivity in identifying patients who would benefit from immunotherapy (p<0.001) and 72% specificity for predicting chemotherapy resistance (p<0.01).In a cohort of 2,500 breast cancer patients, a deep learning algorithm achieved 80% accuracy in predicting complete response to neoadjuvant chemotherapy (p<0.05).AI models incorporating genomic and radiological data showed 68% improvement in predicting treatment failure compared to traditional clinical decision-making (p<0.01).Conclusions: AI-based predictive models show significant promise in predicting treatment responses across a variety of cancer therapies, providing higher sensitivity and specificity than traditional methods.Digital therapeutics further enhance treatment outcomes by improving adherence and alleviating cancer-related psychological distress.These findings highlight the potential of AI and DTx to revolutionize personalized cancer care, optimizing treatment selection and improving patient quality of life.

Bookmark

View Full Paper