Diagnostics

Background/Objectives: Colorectal cancer (CRC) is one of the leading causes of cancer-related mortality worldwide, and accurate histopathological tissue segmentation is critical for timely and reliable diagnosis. Healthcare systems represent complex adaptive environments where diagnostic tools must function reliably across heterogeneous clinical settings, varying staining protocols, and resource-constrained infrastructures. However, existing deep learning segmentation models often require substantial computational resources, limiting their deployment in such settings. This study proposes a novel, resource-efficient colorectal histopathology segmentation network (RCHS-Net) designed for robust clinical deployment across diverse and resource-constrained healthcare environments. Methods: RCHS-Net employs a compact multi-scale encoder with channel recalibration blocks, a gland context module (GCM) with three parallel atrous convolutions and lightweight self-attention for multi-scale contextual feature extraction, and a feature pyramid decoder (FPD) for fine-grained spatial reconstruction. To address the demands of real-world healthcare systems, feature-wise linear modulation (FiLM) conditioning enables class-aware segmentation across multiple tissue categories, while MixStyle augmentation improves stain domain generalization across heterogeneous laboratory and scanner conditions. Results: The model was evaluated on two publicly available benchmark datasets: the EBHI-Seg dataset and the GlaS dataset. On EBHI-Seg, RCHS-Net achieved a mean Dice coefficient of 95.20% and a mean IoU of 91.10% across six colorectal tissue classes, with only 243,226 trainable parameters. On the GlaS benchmark, RCHS-Net attained a Dice score of 93.39% and an IoU of 88.32%, outperforming state-of-the-art methods. Conclusions: RCHS-Net demonstrates that high-accuracy histopathology segmentation can be achieved with a compact architecture, offering a scalable and practical solution for AI-assisted cancer diagnosis across the complex, heterogeneous conditions of real-world healthcare systems, supporting scalable and equitable cancer diagnostics globally.

Background/Objectives: Multisensory inputs generate a common gravity-reference frame, but just the otoliths sense the gravity vector. Graviception is frequently assessed by setting a luminous line to the subjective visual vertical (S.V.V.). Population aging and diabetes prevalence, with insufficient physical activity, imply the need to ponder these factors in clinical assessments. This study aimed to assess S.V.V. accuracy/precision in adults with/without diabetes, according to age, physical activity, and general characteristics. Methods: Participants were 262 adults without diabetes (21–80 years old (y.o.)) and 187 adults with diabetes (28–80 y.o.; matched with 187 without diabetes). All participants had no history of otology/vestibular/neurology/autoimmune/orthopedic/severe renal disease or proliferative retinopathy or traumatic injury or balance complaints. After audiology–vestibular evaluations, the International Physical Activity Questionnaire was self-administered and the S.V.V. was estimated during static and on-axis rotation conditions. Results: In participants without diabetes, S.V.V. precision but not S.V.V. accuracy decreased after the age of 50 years, with no further decrease after the age of 70 years (up to 80 y.o.); the main cofactors contributing to the variability on the S.V.V. precision were physical activity and sitting time, with inconsistent contribution from a history of COVID-19 (R = 0.44 static and 0.35 on-axis, p < 0.00001). In participants with diabetes, the major contribution to the S.V.V. precision variability was from diabetes and age (R = 0.33, static and 31 on-axis, p < 0.00001). Conclusions: Apart from aging, the sensorimotor process affected by insufficient physical activity and sedentary behavior includes vestibular decline. The complex pathophysiology of diabetes may account for the contribution from these cofactors.

Background/Objectives: The first permanent molar (M1) is critical for occlusal development but is highly susceptible to caries and molar–incisor hypomineralization (MIH). When M1 prognosis is poor, extraction may be necessary, requiring accurate evaluation for post-extraction space management. This study aims to develop and validate an automated deep learning framework using YOLOv8n with oriented bounding boxes (OBB) to predict the likelihood of spontaneous space closure following M1 extractions, thereby reducing diagnostic subjectivity and inter-observer variability. Methods: A dataset of 200 pediatric panoramic radiographs was segmented into quadrants and annotated for second permanent molars (M2s) and third molars (M3s). The YOLOv8n-OBB architecture was trained on 640 × 640 pixel images over 100 epochs. The framework integrated M3 presence, M2 Demirjian developmental maturity (proxied by bounding box height), and M2 angulation (via rotation vectors) to map inputs onto an evidence-based clinical decision matrix for prognostic stratification. Results: The model achieved exceptional detection and localization performance with an overall mean average precision (mAP@0.5) of 0.983. Class-specific validation showed high accuracy for M2 (F1-score = 0.978) and M3 (F1-score = 0.904). Quantitative cross-referencing confirmed a seamless mapping of spatial coordinates onto clinical success classes without error propagation. Conclusions: These findings substantiate the YOLOv8n-OBB model as a robust and interpretable decision-support tool. By standardizing prognostic assessments and optimizing treatment planning workflows, the framework serves as an effective aid in pediatric dentistry for managing M1 extractions.