This Indonesian healthcare network operated 22 hospitals and 45 satellite clinics across Java, Sumatra, and Sulawesi, serving over 3.2 million patient visits annually. Diagnostic imaging was a critical bottleneck: the network processed approximately 1.8 million imaging studies per year (X-rays, CT scans, ultrasounds, and MRIs) but had only 38 radiologists across the entire network — a ratio of one radiologist per 47,000 studies.
The radiologist shortage meant that diagnostic turnaround times averaged 72 hours for non-emergency imaging, with some satellite clinics waiting up to five days. Patients in remote clinics often had to travel to urban hospitals for timely diagnosis, creating access barriers that disproportionately affected lower-income populations. Emergency imaging was prioritized but still averaged 4 hours for preliminary reads, during which clinicians made treatment decisions based on their own interpretation rather than specialist radiology review.
A clinical audit revealed a 14% significant discrepancy rate between preliminary clinical reads and final radiologist reports — meaning roughly 1 in 7 imaging studies had clinically meaningful findings that were initially missed or misinterpreted. The network's medical director considered this an unacceptable patient safety risk but could not recruit additional radiologists quickly enough. Indonesia produces fewer than 100 new radiologists annually for a population of 280 million.
Pertama Partners initiated the engagement with an AI Readiness Audit that assessed the network's PACS (Picture Archiving and Communication System) infrastructure, imaging data quality, and radiologist workflow. We analyzed over 320,000 historical imaging studies with paired radiologist reports to build a comprehensive labeled dataset spanning 42 common pathological findings across chest X-ray, abdominal CT, musculoskeletal X-ray, and head CT categories.
Our AI Pilot Program deployed a diagnostic decision support system at five hospitals and eight satellite clinics. The system performed automated preliminary reads of imaging studies, generating structured reports highlighting potential abnormalities with confidence scores, measurements, and comparison to relevant prior studies when available. The model was specifically calibrated for the Indonesian patient population, accounting for the higher prevalence of tuberculosis, dengue-related findings, and tropical infections that Western-trained models underperform on.
The AI was explicitly designed as a triage and decision support tool — not a replacement for radiologist review. Studies flagged as abnormal were prioritized in the radiologist's worklist, enabling critical findings to be confirmed within hours rather than days. Normal studies with high AI confidence were deprioritized but still reviewed. Team Training prepared radiologists to use the AI as a second reader and trained clinic physicians to understand AI confidence levels. Executive Training established governance protocols for AI-assisted diagnostics aligned with Indonesia's Ministry of Health guidelines.
"Indonesia has 280 million people and fewer than 2,000 radiologists. Pertama Partners did not try to replace our radiologists — they multiplied their impact so that a patient in a rural Sumatran clinic gets the same diagnostic quality as someone in Jakarta."— Dr. Siti Nurhaliza, Group Chief Medical Officer
Every transformation starts with a conversation. Let's discuss your challenges and opportunities.
Discuss Your Challenge