Radioiodine‐refractory differentiated thyroid cancer, defined by loss of responsiveness to ¹³¹I therapy or by heterogeneous iodine uptake, accounts for the majority of thyroid cancer‐related deaths; 10‐year survival is approximately 10% with distant metastases. Its definition is not uniform: the 2014 expert consensus and the 2015 American Thyroid Association guideline diverge on heterogeneous uptake and on progression despite substantial uptake, and the 2019 Martinique principles reframe these as risk‐stratification triggers rather than absolute disqualifiers. Refractoriness reflects mitogen‐activated protein kinase‐driven silencing of the sodium iodide symporter; TERT promoter mutations, often co‐occurring with BRAF V600E, are associated with dedifferentiation. The therapeutic landscape has changed over the past decade: antiangiogenic multikinase inhibitors (lenvatinib, sorafenib, and cabozantinib) are established first‐ and second‐line options; molecular profiling supports genotype‐driven and tumor‐agnostic treatment of RET, NTRK, and ALK fusions and BRAF V600E mutations; redifferentiation with MEK or combined BRAF/MEK inhibition can restore radioiodine uptake in selected patients, with responses that vary by molecular subtype; and immune checkpoint inhibitors benefit the rare microsatellite instability‐high or mismatch repair‐deficient subset. Most syntheses organize these options by molecular driver and evidence level, whereas the patient's experience of therapy (toxicity, disruption of daily life, decisional preferences) is acknowledged but seldom carried into recommendations. This review compares those definitions, appraises the evidence for each therapeutic class, defines where ¹²⁴I lesional dosimetry can refine treatment timing, and adds the patient dimension as an explicit third axis, operationalized through two tables: a patient‐centered agent comparison and a scenario‐specific decision matrix. The evidence supports, without yet establishing, early broad molecular profiling, a selective inhibitor where an actionable alteration is present, a multikinase backbone otherwise, and redifferentiation only where tested. Randomized evidence is distinguished throughout from expert opinion. Head‐to‐head comparisons and sequencing data remain limited; circulating tumor DNA and imaging‐derived biomarkers, so far supported only by small cohorts, and adaptive trial designs are the most plausible routes forward. By making treatment burden as explicit as genotype and evidence level, the framework is intended to make treatment selection and shared decision‐making more reproducible.

The accurate segmentation of medical images is crucial to medical care and research; however, many efficient supervised image segmentation methods require sufficient pixel level labels. Such requirement is difficult to meet in practice and even impossible in some cases, e.g., rare Pathoma images. Inspired by traditional unsupervised methods, we propose a novel Chan-Vese model based on the Markov chain for unsupervised medical image segmentation. It combines local information brought by superpixels with the global difference between the target tissue and the background. Based on the Chan-Vese model, we utilize weight maps generated by the Markov chain to model and solve the segmentation problem iteratively using the min-cut algorithm at the superpixel level. Our method exploits abundant boundary and local region information in segmentation and thus can handle images with intensity inhomogeneity and object sparsity. In our method, users gain the power of fine-tuning parameters to achieve satisfactory results for each segmentation. By contrast, the result from deep learning based methods is rigid. The performance of our method is assessed by using four Computerized Tomography (CT) datasets. Experimental results show that the proposed method outperforms traditional unsupervised segmentation techniques.