RESUMO
BACKGROUND: Artificial intelligence (AI) has become a prominent technique for medical diagnosis and represents an essential role in detecting brain tumors. Although AI-based models are widely used in brain lesion segmentation (BLS), understanding their effectiveness is challenging due to their complexity and diversity. Several reviews on brain tumor segmentation are available, but none of them describe a link between the threats due to risk-of-bias (RoB) in AI and its architectures. In our review, we focused on linking RoB and different AI-based architectural Cluster in popular DL framework. Further, due to variance in these designs and input data types in medical imaging, it is necessary to present a narrative review considering all facets of BLS. APPROACH: The proposed study uses a PRISMA strategy based on 75 relevant studies found by searching PubMed, Scopus, and Google Scholar. Based on the architectural evolution, DL studies were subsequently categorized into four classes: convolutional neural network (CNN)-based, encoder-decoder (ED)-based, transfer learning (TL)-based, and hybrid DL (HDL)-based architectures. These studies were then analyzed considering 32 AI attributes, with clusters including AI architecture, imaging modalities, hyper-parameters, performance evaluation metrics, and clinical evaluation. Then, after these studies were scored for all attributes, a composite score was computed, normalized, and ranked. Thereafter, a bias cutoff (AP(ai)Bias 1.0, AtheroPoint, Roseville, CA, USA) was established to detect low-, moderate- and high-bias studies. CONCLUSION: The four classes of architectures, from best-to worst-performing, are TL > ED > CNN > HDL. ED-based models had the lowest AI bias for BLS. This study presents a set of three primary and six secondary recommendations for lowering the RoB.
RESUMO
BACKGROUND AND PURPOSE: The Head and Neck Imaging Reporting and Data System (NI-RADS) surveillance template for head and neck cancer includes a numeric assessment of suspicion for recurrence (1-4) for the primary site and neck. Category 1 indicates no evidence of recurrence; category 2, low suspicion of recurrence; category 3, high suspicion of recurrence; and category 4, known recurrence. Our purpose was to evaluate the performance of the NI-RADS scoring system to predict local and regional disease recurrence or persistence. MATERIALS AND METHODS: This study was classified as a quality-improvement project by the institutional review board. A retrospective database search yielded 500 consecutive cases interpreted using the NI-RADS template. Cases without a numeric score, non-squamous cell carcinoma primary tumors, and primary squamous cell carcinoma outside the head and neck were excluded. The electronic medical record was reviewed to determine the subsequent management, pathology results, and outcome of clinical and radiologic follow-up. RESULTS: A total of 318 scans and 618 targets (314 primary targets and 304 nodal targets) met the inclusion criteria. Among the 618 targets, 85.4% were scored NI-RADS 1; 9.4% were scored NI-RADS 2; and 5.2% were scored NI-RADS 3. The rates of positive disease were 3.79%, 17.2%, and 59.4% for each NI-RADS category, respectively. Univariate association analysis demonstrated a strong association between the NI-RADS score and ultimate disease recurrence, with P < .001 for primary and regional sites. CONCLUSIONS: The baseline performance of NI-RADS was good, demonstrating significant discrimination among the categories 1-3 for predicting disease.
