Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.
Researchers developed and validated machine learning models to predict posttraumatic epilepsy onset at 2, 5, and 10 years after first TBI documentation in 107,987 post-9/11 US veterans, using only routine preinjury clinical data up to the month of injury. An optimized random forest achieved AUCs of 0.75 to 0.73 across horizons on held-out test data and enabled high-risk stratification.
Researchers compared two ways to allocate scarce Medicaid care-management phone outreach each month for 164,063 beneficiaries in Washington and Virginia. Using a causal forest to estimate individualized treatment effects, they found targeting the top decile by predicted effect prevented 13.3 acute events per 2000 members per month, compared with 2.5 events under conventional top-decile risk targeting.
By August 2025, a scoping review of 73 studies published between 2015 and 2024 examined AI in sports biomechanics, focusing on wearable technology, motion analysis, and injury prevention. It reported that convolutional neural networks reached 94% agreement with experts, computer vision was within 15 mm of marker-based systems, and integrated AI systems were associated with a 23% reduction in reinjury rates.
On 2025-10-04, a peer-reviewed review in Foods synthesized 25 studies selected from 124 Scopus records from 2005-2025 to map machine learning use for quality control in food production. It organized findings into six domains covering quality applications, defect detection and visual inspection, ingredient optimization, packaging sensors and predictive QC, supply chain traceability, and Industry 4.0 models.
Between 2023 and 2025, educators at a single academic institution integrated an 8-hour interactive AI rotation into diagnostic radiology residency, combining didactics with lab modules on convolution, radiomics, machine learning, deep learning, evaluation, bias, and clinical cases for 27 residents across three cohorts.
A TRIPOD+AI scoping review of 61 studies examined AI and radiomics applied to CT, MRI, and FDG-PET/CT for histologically confirmed oropharyngeal squamous cell carcinoma, focusing on HPV status prediction and treatment deintensification. CT was the dominant modality, manual segmentation and handcrafted radiomics with machine learning were most common, and HPV models reported AUCs from 0.65 to 0.95.
Researchers analyzed 664 U.S. occupations across 128 skill dimensions, using K-means clustering to identify five distinct groups and scoring each for AI Exposure via AIOE and for protective job features via PBI, with PCA to examine how exposure and buffering relate to skill composition.
A peer-reviewed study tested the Mistral-NeMo language model on 1000 operative notes from 177 providers to verify Current Procedural Terminology codes for lower-extremity orthopaedic surgery. When CPT billing descriptions were included in the prompt, the model correctly identified 90% of true codes and rejected 99.80% of incorrect codes.
ace