AI Caries Detection: Faster, But Not Smarter

AI Caries Detection: Faster, But Not Smarter

your software company is pitching AI caries detection. It's faster than visual inspection. It catches more lesions. It's backed by a study showing 94% accuracy on controlled images.

you're thinking about buying it. Here's why you shouldn't - not yet, anyway.

OPERATOR MATH

Let's price out the AI caries software vs. your current workflow.

Current state (no AI): You catch 88% of carious lesions on visual X-ray inspection. Your false positive rate: 6% (you flag something as suspicious, do exploratory work, find it's normal anatomy). Annual X-rays: 1,600 patients × 2 X-rays = 3,200 images reviewed. Time per image review: 45 seconds. Total annual time: 2,400 minutes = 40 hours. Your hourly rate (calculated): $180/hour. Cost of your time: $7,200/year.

AI scenario: Software subscription: $5,500/year. AI catches 92% of lesions (4% improvement). But false positive rate: 14%. That's 3,200 × 14% = 448 false flags annually. You spend 2 minutes per false flag evaluating it (compared to 45 seconds without AI). Extra time: 448 × 2 min = 896 minutes = 15 hours. Cost: 15 hours × $180 = $2,700 in extra cognitive load. Total cost: $5,500 (software) + $2,700 (your time) = $8,200/year.

The gain: You catch 4% more lesions. On 3,200 X-rays, that's 128 additional lesions flagged. Assume 40% are real and lead to treatment (the rest are early lesions you monitor). That's 51 additional treatments/year. Average treatment value: $220 (small filling). Incremental production: 51 × $220 = $11,220. After 35% overhead: $7,293 in net production gain.

Net financial outcome: $7,293 (gain) - $8,200 (AI cost) = -$907 annual loss. You're paying $907/year for marginal clinical improvement that doesn't cover its cost. The DSO math works because they're starting from 30% detection (burned-out associates). Your 88% baseline makes AI uneconomical.

THE TAKEAWAY

Skip AI caries detection until one of these conditions changes:

1. Wait for pricing to drop below $1,200/year. At that threshold, even marginal clinical gains break even financially. Check vendor pricing annually - competition will drive this down by 2027-2028.

2. Track your current detection rate. If you're honestly catching <80% of lesions (audit your X-rays against a specialist's review), AI might be worth it. But most competent general dentists are at 85-90% already.

3. Monitor false positive improvements. When vendor data shows real-world specificity >95% (not lab studies - actual clinic deployments), revisit the decision. That's the threshold where cognitive overhead becomes manageable.

4. If you run a DSO or multi-location practice with associate variability, the math changes. AI creates consistency across providers. A 10-location DSO might justify $55K/year in software if it standardizes diagnosis and reduces liability from missed lesions.

5. Focus your money on better X-ray equipment instead. A $15K investment in a high-resolution digital sensor improves image quality permanently and helps YOU see lesions better - no subscription, no false positives, no cognitive overhead.