Singapore's academic institutions are at the forefront of harnessing artificial intelligence (AI) to revolutionize cancer diagnosis, addressing a pressing public health challenge. Cancer is the leading cause of death in the country, accounting for nearly 25% of all deaths in 2023, with 91,574 cases reported between 2019 and 2023—roughly 18,000 new diagnoses each year. Notably, diagnoses among those under 40 have surged by 34% from 2003-2007 to 2019-2023, underscoring the urgency for faster, more accurate detection methods. Universities like the National University of Singapore (NUS) and Duke-NUS Medical School are driving research that integrates AI into imaging, pathology, and blood tests, accelerating detection and improving outcomes for common cancers such as breast, liver, and lung.
🤖 The Rise of AI in Cancer Detection: From Global Benchmarks to Singapore Innovations
Artificial intelligence refers to computer systems that perform tasks requiring human intelligence, such as pattern recognition in medical images. In cancer diagnosis, AI—particularly deep learning models like convolutional neural networks (CNNs)—analyzes mammograms, CT scans, and biopsies faster and often more accurately than humans alone. Globally, studies like the RADIOHEAD prospective cohort, published in 2025, demonstrate machine learning's prowess in predicting treatment responses in advanced lung cancer using multi-omic data from over 500 patients, detecting immunotherapy responses up to five months earlier via blood tests. Featured in regional outlets like BioSpectrum Asia, such findings inspire Singapore's push toward AI-accelerated diagnostics tailored to Asian populations with higher rates of dense breasts and liver cancers.
Singapore's ecosystem, bolstered by initiatives like AI Singapore, fosters university-led spin-offs and collaborations with hospitals such as National University Hospital (NUH) and Singapore General Hospital (SGH). This synergy positions the nation as an APAC hub for oncology AI, with the market projected to grow from $41.7 million in 2023 to $247.4 million by 2030.
🔬 NUS-Led Breakthrough: AI Boosts Breast Cancer Detection in Mammography
Breast cancer is among Singapore's top cancers, with dense breast tissue in many Asian women complicating detection. A landmark 2025 study from NUS researchers at NUH evaluated FxMammo, an AI tool developed by FathomX—a NUS/NUHS spin-off approved by Singapore's Health Sciences Authority (HSA). Led by Dr. Serene Si Ning Goh from NUS Yong Loo Lin School of Medicine and Saw Swee Hock School of Public Health, the multi-reader multi-case trial involved 17 radiologists (consultants, senior/junior residents) reviewing 500 de-identified mammograms (250 malignant, 250 benign).
The process worked as follows: Radiologists first read without AI, then with FxMammo's heatmaps and risk scores after a one-month washout. AI standalone achieved an area under the receiver operating characteristic curve (AUROC) of 0.93—outperforming juniors (0.84) and seniors (0.85), nearing consultants (0.90). With AI assistance:
- Junior residents' sensitivity rose from 56.9% to 61.6% (P<0.001), specificity from 94.6% to 96.3% (P=0.02).
- Senior residents saw accuracy jump from 76.1% to 80.4% (P=0.002).
- Time savings averaged 18 seconds per non-malignant read, benefiting overburdened juniors most (19.7s overall).
- Greatest gains in dense breasts (66-68% of cases), where AUROC improved from 0.82 to 0.84.
Inter-reader agreement strengthened (kappa 0.54-0.60), highlighting AI's role in standardizing reads amid Singapore's radiologist shortages.Read the full JMIR study.
👥 Radiologists' Perspectives: Building Trust in AI Tools
A companion NUS study in Cancers journal surveyed 17 experienced radiologists and interviewed 10, revealing nuanced views on AI integration into Singapore's double-reader screening. While 64.7% endorsed AI as a companion, only 29.4% saw it matching human performance; 76.5% rejected full replacement.
Confidence hinged on local validation (9.3/10) and guidelines (8.8/10). Benefits included triage (41.2%) and consistency in routine cases, but challenges like false positives increasing workload (58.8%) and medico-legal concerns loomed. Experts advocated user-friendly interfaces, training, and human oversight for discordant cases—key for adoption in dense-breast prevalent populations.
🩸 Beyond Imaging: Fragle AI for Blood-Based Cancer Tracking
Shifting to liquid biopsies, A*STAR's Genome Institute of Singapore (GIS), collaborating with NUS-linked National Cancer Centre Singapore (NCCS), unveiled Fragle in Nature Biomedical Engineering (March 2025). Led by Dr. Anders Skanderup, Fragle analyzes circulating tumor DNA (ctDNA) fragment sizes from small blood samples to detect minimal residual disease (MRD) and monitor treatment—faster and cheaper (<SGD$50 vs. $1000+).
Unlike mutation-specific methods, Fragle's versatility suits diverse cancers, showing consistency across samples. Ongoing NCCS trials track lung cancer patients bi-monthly, spotting relapses pre-scan. This aligns with Singapore's precision oncology push.
肝 Liver Cancer Advances: TIMES Score and Project RAPIER
Hepatocellular carcinoma (HCC), linked to hepatitis/NAFLD, claims many lives. SGH researchers' TIMES score (Nature, July 2025) predicts recurrence post-resection with 82% accuracy by analyzing immune cell spatial patterns via AI—superior to clinicopathologic factors. Validation at SGH/NCCS continues.
Project RAPIER, pooling 5,000+ SGH/NCCS images/pathology for liver lesions, builds AI for automated labeling/reporting, reducing turnaround and errors.
- Identifies benign vs. malignant lesions precisely.
- Integrates radiology-pathology data lakes for training.
- Aims at real-time clinical use.
🌍 Broader University Contributions and Spin-Offs
Duke-NUS advances AI for laryngeal screening in low-resource settings and global diagnostics. NUS reviews AI for breast cancer management, from detection to prognosis. Startups like Qritive (prostate AI, HSA-approved) stem from this ecosystem, aiding pathologists with Gleason grading.
Stakeholders praise workload reduction (up to 70%) and access gains, but stress validation.
⚠️ Challenges: False Positives, Ethics, and Integration
Despite promise, hurdles persist: AI false positives inflate workloads; bias in datasets risks inequities; regulations lag. Singapore addresses via HSA approvals, ethics frameworks from NUS Centre for Biomedical Ethics. Solutions include hybrid human-AI workflows and diverse training data.
🚀 Future Outlook: Precision Medicine and Careers
By 2030, AI could halve diagnostic times, boosting survival. Singapore's RISE institute eyes preventive tools. For aspiring researchers, opportunities abound in AI-oncology at NUS/Duke-NUS—explore research jobs, faculty positions, or career advice on AcademicJobs.com. Internships in data science for health prepare for postdocs via postdoc jobs.
Real-world cases: FxMammo cut juniors' errors; Fragle flags early relapses. Implications span policy (national guidelines) to patients (earlier interventions). As Prof. Gerald Koh (NUS) notes, collaborative validation is key.
