Smartphone application based on artificial intelligence to characterize the shape of the stool

In a recent study published in the American Journal of Gastroenterologyresearchers at Cedars-Sinai Medical Center in the United States evaluated an artificial intelligence (AI)-based smartphone application trained to assess the characteristics of a patient’s stool.

Study: Smartphone app using artificial intelligence is superior to subject’s self-report when assessing stool shape. Image Credit: Josep Suria / Shutterstock

Background

Functional gastrointestinal (GI) disorders, especially luminal disorders, require a patient to self-report stool form and frequency. However, since diarrhea symptoms common in patients with irritable bowel syndrome with diarrhea (IBS-D) are subjective, the inability to accurately report or assess stool form and frequency makes it difficult to determination of the effectiveness of therapeutic interventions in these conditions.

The Bristol Stool Scale (BSS) is a 7-point scale approved by the United States Food and Drug Administration (US-FDA) that ranks stool consistency from 1 (hard lumps) to 7 (liquid). However, inconsistent and inaccurate self-reporting of stool forms causes problems, especially among IBS-D cases. In such cases, AI algorithms could help systematically evaluate digital images of a person’s stool.

About the study

In the present study, researchers recruited subjects participating in a randomized clinical trial for IBS-D to validate AI determinations for stool images based on five distinct visual characteristics of stool, namely, edge blurring , consistency, BSS, volume and fragmentation. In another group of individuals in the same trial, they assessed how well app results aligned with self-reported BSS scores. Finally, the team compared the subject-determined and AI-determined stool characteristic scores with the standardized diarrhea severity scores.

Participating subjects captured all stool images during the two-week screening phase of the trial. The app processed the results and determined five visual stool characteristics and bowel frequency. Two experts validated the AI ​​images of the first third of the subjects. Later, the team also classified the AI-annotated stool images, self-reported by study participants and two experts, into categories, BSS 5 (diarrhea). Finally, the team calculated the sensitivity, specificity, accuracy, and diagnostic odds ratios of self-reported, AI-scored BSS scores by comparing them to expert assessments, which they considered the gold standard. .

Study results

There were a total of 39 study participants, 14 of whom provided 219 stool images for the validation phase. The team used data from the other 25 subjects for the implementation phase. The RN and expert gastroenterologists presented BSS scores from one to seven and their ratings were in good agreement for all five stool characteristics, as were the RN and expert ratings.

The mean rates of specificity and sensitivity of AI-graded BSS categorization were 11% and 16% higher, respectively. The mean odds ratio and diagnostic accuracy rate were higher for AI at 30.64 vs. 3.67 and 95% vs. 89% when compared to subject-reported scores. The agreement between the BSS scores reported by the subjects and those scored by the AI ​​was 0.31 during the validation phase, but reached a value of 0.61 during the implementation phase. On average, the visual characteristics of stool determined by AI between the two phases remained similar.

In addition, the authors observed a good correlation between mean daily BSS scores noted by AI and diarrhea severity scores in SII-D subjects. The other four app-reported visual stool characteristics also correlated quite well with diarrhea severity scores. Notably, all subjects found the app easy to use, and 50% of those who responded to user experience questions described their experience as easy and very enjoyable.

conclusion

Previously, IBS drug testing often depended on weekly assessments of gastrointestinal symptoms. Later, the US FDA developed new guidelines for IBS, which required trial sponsors to ask all participants to report and characterize symptoms daily to improve accuracy. Yet, the BSS remains essential for assessing self-reported stool hardness and assessing individual stool types during clinical trials.

Inaccurate self-reporting of stool shape could stem from inadequate understanding of the topic and recall bias. Although intuitive, the patient must be familiar with the BSS to avoid any misperceptions. This becomes difficult when subjects with diarrhea report an average daily BSS despite having multiple and varied bowel movements in a day. The results of the current study confirm that the self-reported daily scores differed from the BSS scores given by the two experts.

The AI ​​catalogs characterized stool shape in an objective “true” sense, as a subject documented each bowel movement in a photo. Digital stool images have enabled a comprehensive assessment of drug effect and objective quantification of side effects of therapies for bowel disorders. Additionally, these images assessed stool characteristics beyond the BSS. The evaluation of four new characteristics made it easier to consider each saddle separately and avoided the need to collect daily averages. Overall, the observed pattern of test characteristics suggested that AI scores were superior. Additionally, they reduced trial costs because sponsors could now design trials with large numbers of subjects to mitigate the effect of inconsistency and inaccuracy of self-reporting. As a result, in the future, objective measurement of stool form would require fewer subjects to test drugs.

To conclude, the AI-based app used in the study accurately characterized stool versus self-report and correlated well with diarrhea severity. It has the potential to become a valuable tool for use in luminal gastrointestinal disease trials, including IBS-D, as it was both accurate and objective in defining stool characteristics beyond BSS. .

Journal reference:

  • Pimentel, Mark, Mathur Ruchi, Wang Jiajing, Chang Christine, Hosseini Ava, Fiorentino Alyson, Rashid Mohamad, Pichetshote Nipaporn, Basseri Benjamin, Treyzon Leo, Chang Bianca, Leite Gabriela, Morales Walter, Weitsman Stacy, Kraus Asaf, Rezaie Ali, A Smartphone app using artificial intelligence is superior to subject’s self-report when assessing stool form, The American Journal of Gastroenterology: July 2022, DOI: 10.14309/ajg.0000000000001723, https://journals.lww.com/ajg/Fulltext/2022/07000/A_Smartphone_Application_Using_Artificial.24.aspx

Comments are closed.