Latest Research
All publications from the Cancer3.AI database, newest first.
Assessing agreement with a single-center expert consensus: artificial intelligence-assisted teleultrasound for thyroid nodules categorized as C-TIRADS 4A or higher.
Yi C, et al
Researchers in China evaluated whether an artificial intelligence system could match expert-level accuracy in assessing suspicious thyroid nodules identified through teleultrasound at community health centers. The study analyzed 80 thyroid nodules classified as C-TIRADS 4A or higher — a Chinese imaging risk classification indicating elevated concern for malignancy — comparing diagnoses made by community clinicians, remote specialist teleultrasound, and an AI diagnostic tool. Community medical institutions showed poor agreement with specialist teleultrasound (kappa = 0.20), while the AI system demonstrated strong agreement with specialists (kappa = 0.80) and achieved a macro-average AUC of 0.92, comparable to expert performance. At the C-TIRADS 4A threshold, the AI system reached a sensitivity of 97.1% and a specificity of 100%, indicating highly reliable detection of potentially malignant nodules. These findings suggest that AI-assisted teleultrasound could help bridge the diagnostic gap between community health centers and specialist institutions, potentially improving early detection and management of thyroid cancer in underserved or rural populations.
Frontiers in endocrinology
Source →Enhanced early detection of thyroid cancer using ensemble machine learning and serum proteomics.
Zhang D, et al
Researchers developed a blood-based diagnostic model for early thyroid cancer detection by combining serum proteomics with ensemble machine learning algorithms. The study analyzed blood samples from 414 thyroid cancer patients and 430 healthy controls using MALDI-TOF mass spectrometry, then trained and validated multiple machine learning models on the resulting protein data. The best-performing integrated model showed excellent accuracy, and interpretability tools (SHAP and LIME) confirmed that its power came from the coordinated contribution of multiple protein features rather than any single marker. A streamlined version of the model, built on just 12 key peptide peaks linked to immune regulation and lipid metabolism, retained high diagnostic accuracy and outperformed single-protein biomarkers in clinical benefit analyses. These findings offer a promising non-invasive strategy for catching thyroid cancer at an early stage, potentially improving outcomes for patients who currently go undiagnosed until the disease has already spread.
Frontiers in oncology
Source →Retrospective study of hospitalized breast cancer patients in a Zhuhai-based hospital: analysis spanning over 20 years.
Fan R, et al
A retrospective study published in Frontiers in Oncology examined 20 years of breast cancer care at a regional hospital in Zhuhai, Southern China, analyzing data from 5,052 patients hospitalized between 2004 and 2024. The research revealed that most patients were diagnosed at an early stage (Stage I, 61.3%), with a mean age of 50.5 years, and that hospital stays shortened dramatically from nearly 17 days to under 5 days over the study period. A notable 406% surge in hospitalizations occurred in 2022, while treatment approaches shifted toward breast-conserving surgery and neoadjuvant chemotherapy, and total treatment costs fell by nearly 59% after peaking in 2019. Adjuvant chemotherapy emerged as the strongest predictor of cure, with patients receiving it nearly six times more likely to achieve a curative outcome. Out-of-pocket expenses also declined from 49.2% to 37.4%, suggesting improved insurance coverage for patients. These findings highlight meaningful progress in breast cancer management in regional Chinese healthcare settings and underscore the importance of chemotherapy in treatment planning.
Frontiers in oncology
Source →Development and validation of a machine learning model for predicting unplanned removal of totally implantable venous access ports in patients with breast cancer.
Wang Y, et al
Researchers from a tertiary hospital in Chongqing, China, conducted a retrospective study involving 1,258 breast cancer patients to identify risk factors for unplanned removal of totally implantable venous access ports (TIVAPs) — small devices surgically placed under the skin to deliver chemotherapy — and to build a machine learning model capable of predicting which patients are at highest risk. Four machine learning algorithms were tested, and the XGBoost model emerged as the top performer, achieving an area under the ROC curve of 0.826 in the training set and 0.751 in the validation set, reflecting strong and reliable predictive accuracy. Key independent risk factors identified included body mass index, cancer stage, the route used for port implantation, number of puncture attempts, blood coagulation indices, neutropenia, and duration of catheter placement. The model was further validated through calibration curves and decision curve analysis, confirming both its statistical soundness and its practical clinical value. These findings offer clinicians a data-driven tool to flag high-risk patients early, enabling targeted preventive measures that could reduce complications, unnecessary port removals, and treatment interruptions for breast cancer patients undergoing chemotherapy.
Frontiers in oncology
Source →Immune Checkpoint Inhibitor Therapy Induced Hypophysitis; A Tertiary Care Centre Experience; Highlighting Detection, Treatment Outcomes and Long Term Endocrinopathies and Recovery.
Mathara Diddhenipothage SA, et al
Researchers at a tertiary care centre reviewed 22 patients who developed immune checkpoint inhibitor-induced hypophysitis (ICI-Hp), a form of pituitary gland inflammation triggered by cancer immunotherapy, to better understand its diagnosis, treatment, and long-term hormonal consequences. The study found that ICI-Hp most commonly appeared around 13 weeks into treatment, typically after the third cycle of combination therapy with ipilimumab and nivolumab, with headache, fatigue, nausea, and low sodium levels being the predominant symptoms. All patients developed secondary adrenal insufficiency, and imaging revealed pituitary gland enlargement in nearly two-thirds of cases, with three patients showing mild compression of the optic chiasm. Importantly, while thyroid and gonadal hormone deficiencies recovered in a substantial proportion of patients, adrenal insufficiency persisted in every patient throughout follow-up, underscoring the need for lifelong glucocorticoid replacement therapy. The findings highlight that oncology teams and patients should be specifically alerted to watch for warning symptoms after the third cycle of combination immunotherapy, enabling earlier detection and management of this potentially serious complication.
Clinical endocrinology
Source →