OpenRounds Editorial

Healthcare AI Daily Briefing

Monday, October 12, 2026

New to OpenRounds? Get the free daily email or see how we work.

Clinical Practice

CP-enabled random forest improves sensitivity for SPMS classification versus Lorscheider criteria.

Preprint

The CP-enabled random forest model was trained on SMSReg-2022 and prospectively validated on SMSReg-2025 (15,719 individuals) to distinguish relapsing-remitting from secondary-progressive multiple sclerosis, achieving higher sensitivity than the Lorscheider criteria (98% vs. 71%) while maintaining good calibration. [1]

Diagnostics & Pathology

Data‑efficient ML achieves 94% accuracy detecting duodenal neoplasia on external test.

Preprint

Using Hibou‑B embeddings, PCA, a Gaussian mixture model and a Random Forest on a five‑hospital dataset of 8,027 normal and 401 neoplastic duodenal biopsy WSIs, the pipeline achieved 94% accuracy on an external test hospital. [2]

Research & Evidence

Smartphone‑based AI estimates proptosis within 2 mm of Hertel in most thyroid eye images.

Journal article

In a multicenter validation of 361 thyroid eye disease patients at Korean and Spanish sites, a deep‑learning model fine‑tuned for nonstudio facial photographs achieved Hertel‑level agreement and reliable severity discrimination across two ethnicities and 12 smartphone models, for both patient self‑acquired selfies and clinician‑taken routine images. [3]

Medical Imaging

Delta Radiomics and Tabular Transformer yield better lumbar disc degeneration risk prediction.

Journal article

In a retrospective study of 663 L4‑L5 MRI subjects, Delta Radiomics features processed by a Tabular Transformer model achieved an AUC of 0.848 for lumbar intervertebral disc degeneration risk assessment, surpassing the linear model’s AUC of 0.804. Decision curve analysis and a Brier score of 0.145 indicated higher net clinical benefit across risk thresholds. [4]

Sources

Prospective validation of an uncertainty-aware machine learning model for secondary progressive multiple sclerosis classification · medRxiv Original source
Data-efficient machine learning for the detection of duodenal neoplasia: a multi-centre study · medRxiv Original source
Smartphone-Based Deep Learning Exophthalmometry in Thyroid Eye Disease: A Multicenter, Multiethnic Validation Study. · Thyroid : official journal of the American Thyroid Association Original source
Digital twin validation and predictive value of delta radiomics based on tabular transformer for synergistic muscle imbalance in lumbar disc degeneration. · European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society Original source

CP-enabled random forest improves sensitivity for SPMS classification versus Lorscheider criteria.

Data‑efficient ML achieves 94% accuracy detecting duodenal neoplasia on external test.

Smartphone‑based AI estimates proptosis within 2 mm of Hertel in most thyroid eye images.

Delta Radiomics and Tabular Transformer yield better lumbar disc degeneration risk prediction.

Sources

Get the free healthcare AI briefing