Clinical Validation
Four peer-reviewed studies. Three top academic medical centers. One AI.
Dieta's stool image AI has been independently validated by gastroenterologists and hepatologists
at Cedars-Sinai, Mayo Clinic, and Virginia Commonwealth University across three indications:
irritable bowel syndrome, ulcerative colitis, and hepatic encephalopathy.
4Peer-reviewed papers
~130Patients studied
3Indications
3Academic medical centers
Medical Center · Los Angeles
Foundational Validation
Irritable Bowel Syndrome (IBS-D)
AI outperforms patient self-report in assessing stool form
Embedded in a randomized controlled drug trial, the Dieta application classified stool images on
five visual characteristics and was compared blindly against two expert gastroenterologists
as the gold standard, then against patient self-report. The AI matched the experts and substantially
outperformed patients themselves.
0.78–0.85ICC vs. expert GIs (BSS)
8.4×Higher diagnostic odds ratio vs. self-report
95%Accuracy vs. 89% self-report
IBD Clinical Research
UC + Inflammation
Acute Severe Ulcerative Colitis (Inpatient)
AI stool characteristics correlate with objective inflammation
In hospitalized patients with acute severe UC, AI-measured Bristol Stool Scale, stool consistency,
and fragmentation significantly correlated with serum C-reactive protein — the
first evidence that the Dieta AI tracks not just patient symptoms but underlying biology.
0.81AUROC predicting CRP from BSS (toilet imgs)
151Stool images analyzed
p<0.05BSS, consistency, fragmentation vs CRP
Richmond VA Medical Center · VCU Health
Behavioral Impact
Hepatic Encephalopathy / Cirrhosis
AI feedback improves patient self-assessment and triggers self-titration
Patients with cirrhosis on lactulose used the app for two phases — with and without AI feedback
communicated back to them. AI feedback improved the agreement between patient and objective stool
assessment by more than 5×, and patients self-titrated their lactulose dose
appropriately without study-team input.
r = 0.72Self-BSS vs. AI-BSS (with feedback)
r = -0.16Self-BSS vs. AI-BSS (no feedback)
AutoPatient self-titrated lactulose dose
GI & Hepatology · Rochester, MN
Active Titration
Hepatic Encephalopathy / Cirrhosis
App-driven lactulose titration improves BSS goal achievement
In a structured 28-day feasibility trial, Mayo Clinic patients received daily lactulose-dosing
recommendations through the Dieta app based on AI-derived Bristol Stool Scale. In the
less-compliant subgroup — the patients pharma cares about most — BSS goal
achievement jumped from 56% to 79% (p = 0.041).
56% → 79%BSS goal achievement (less-compliant)
95%Would recommend the app
73.6%Enrollment rate