AI May Improve Breast Mitotic Scoring

Junior pathologists showed higher mitotic score accuracy and agreement with artificial intelligence assistance in a preliminary single-center reader study

By
Andrea Surnit
May 13, 2026
4 min

MDSpire News

At a Glance

Category	Detail
Condition	Breast Carcinoma Mitotic Scoring
Key Mechanisms	AI-assisted detection of mitoses and identification of mitotic hotspots to support scoring according to the Elston and Ellis grading system.
Target Population	Junior pathologists evaluating breast carcinoma specimens.
Care Setting	Single-center study at Bicêtre Hospital.

Key Highlights

AI assistance improved mitotic score accuracy from 62% to 76% for one investigator and from 64% to 78% for another.
Weighted Cohen’s kappa for agreement with expert consensus increased significantly with AI assistance.
AI improved consistency in hotspot selection, with intersecting regions increasing from 46% to 80% for one investigator.
Accuracy gains were most notable in diagnostically challenging subgroups, particularly mitotic score 2 and 3 cases.
The study emphasizes the variability in expert mitotic scoring and the potential of AI to enhance diagnostic reliability.

Guideline-Based Recommendations

Diagnosis

Utilize AI tools to assist in mitotic scoring to improve accuracy and reproducibility.

Management

Incorporate AI-assisted scoring in pathology workflows to enhance diagnostic confidence.

Monitoring & Follow-up

Regularly evaluate the performance of AI tools against expert consensus to ensure reliability.

Risks

Caution is advised due to the study's limited scope involving only two junior pathologists and retrospective specimen evaluation.

Patient & Prescribing Data

Patients with breast carcinoma undergoing mitotic scoring.

AI tools may aid pathologists in more accurately assessing tumor aggressiveness based on mitotic activity.

Clinical Best Practices

Encourage pathologists to use AI tools as adjuncts rather than replacements for their judgment.
Implement structured training for pathologists on the use of AI-assisted scoring systems.
Consider the variability in expert scoring when interpreting AI-assisted results.