Takeji Saitoh

Background: Cytology artificial intelligence (AI) systems often depend on whole‐slide images (WSIs) or high‐magnification cell images and are commonly optimized for a single specimen type. We developed CYTOLONE‐v2, a contrastive language‐image pretraining (CLIP)–based hierarchical classification framework for cervical and urinary cytology that is designed for individual‐field inference without WSI input. Methods: CYTOLONE‐v1.1, previously trained for cervical cytology, was further trained by low‐rank adaptation using WSI‐derived urinary cytology fields equivalent to ×10 magnification. The 90‐case source cohort was divided at the case level into training (72 cases; 360 fields) and validation (18 cases; 90 fields) subsets. Five diagnosis‐representative fields were manually selected from each case. Field‐level performance was evaluated in an independent urinary test cohort of 40 cases (20 Reactive change and 20 Urothelial carcinoma; 200 fields), and retention was assessed using an independent cervical cytology test set. Results: In urinary cytology, CYTOLONE‐v2 achieved field‐level accuracies of 64.0% for Anomaly and Malignancy and 72.5% for Diagnosis. At the Diagnosis level, sensitivity for urothelial carcinoma was 99.0% (95% confidence interval [CI], 96.7%–100%), whereas specificity against Reactive change was 46.0% (95% CI, 30.0%–62.0%). Cervical cytology accuracy was retained at the Anomaly level (94.8% vs. 94.7% with CYTOLONE‐v1.1), decreased modestly at the Malignancy level (91.2% vs. 93.5%), and decreased more at the Bethesda and Diagnosis levels. Conclusions: CYTOLONE‐v2 demonstrates the technical feasibility of extending a hierarchical cytology classification framework from cervical to urinary cytology. The findings are limited to selected WSI‐derived fields and a Reactive change‐versus‐Urothelial carcinoma test cohort; external, case‐level, direct‐microscopy and prospective workflow validation is required before clinical utility can be established.

Background/Objectives: Transthoracic echocardiography (TTE) is a non-invasive tool for real-time assessment of cardiac motion and blood flow. It is widely used in emergency and bedside settings as a Focus Cardiac Ultrasound (FoCUS) device. However, standardized training methods and adequate educational environments are limited. Methods: A TTE image assessment artificial intelligence (AI) system was developed in this study, focusing on probe positioning and image quality for non-supervised TTE practice. Results: The view classification model achieved a high F1-score of 0.956. The position evaluation model achieved F1-scores of 0.678, 0.864, and 0.831 for the parasternal long-axis, parasternal short-axis, and apical four-chamber views, respectively. The quality evaluation model achieved F1-scores of 0.674, 0.845, and 0.746. Combining the position and quality models improved the F1-score for the parasternal long-axis view to 0.714, showing the benefit of integrating views with lower baseline performance. Conclusions: This study presents a novel AI-based educational system that assesses probe position and image quality in TTE. The model was developed using a custom dataset of healthy young adults that reflects beginner-level training scenarios, including many suboptimal images similar to those commonly acquired by novices. The proposed framework, which integrates position and quality models, lays the groundwork for future AI-assisted ultrasound training, particularly in unsupervised or resource-limited settings.

Background: ECG interpretation is sometimes difficult due to baseline fluctuations and electrode detachments when placed on the subjects’ front side, leading to misinterpretation of the rhythms and phases of the cardiac cycle. We aimed to compare the differences in the wave amplitudes and respiratory variations between conventional ECG electrode positioning on the front side of patients and an alternative position on the backs of patients. Methods: Echocardiography was performed in 85 patients lying in the left lateral position. We attached the red electrode to the right clavicle, the yellow to the left clavicle, and the green to the left lateral abdomen on the front side of the patients; on the back, we attached the electrode to the right clavicle, the right upper posterior iliac spine, and the left upper posterior iliac spine. Results: The ECG monitor amplitudes were greater on the front side compared to the back side, but the BF-breath values were smaller on the back side (6.0 pixels) compared to the front side (10.5 pixels, p<0.05). The P wave amplitude divided by the BF-breath on the back side was greater than that seen on the front side (2.8 vs. 1.8, p<0.05), whereas the QRS amplitude divided by the BF-breath was 15.0 and 16.3, respectively (p=ns). Conclusion: As an alternative to front-side ECG monitoring, electrodes placed on the back can help avoid misinterpretation of the ECG rhythms and the phases of the cardiac cycle due to respiration during echocardiography.

In local and global disaster scenes, rapid recognition of victims’ breathing is vital. It is unclear whether the footage transmitted from small drones can enable medical providers to detect breathing. This study investigated the ability of small drones to evaluate breathing correctly after landing on victims’ bodies and hovering over them. We enrolled 46 medical workers in this prospective, randomized, crossover study. The participants were provided with envelopes, from which they were asked to pull four notes sequentially and follow the written instructions (“breathing” and “no breathing”). After they lied on the ground in the supine position, a drone was landed on their abdomen, subsequently hovering over them. Two evaluators were asked to determine whether the participant had followed the “breathing” or “no breathing” instruction based on the real-time footage transmitted from the drone camera. The same experiment was performed while the participant was in the prone position. If both evaluators were able to determine the participant’s breathing status correctly, the results were tagged as “correct.” All experiments were successfully performed. Breathing was correctly determined in all 46 participants (100%) when the drone was landed on the abdomen and in 19 participants when the drone hovered over them while they were in the supine position (p < 0.01). In the prone position, breathing was correctly determined in 44 participants when the drone was landed on the abdomen and in 10 participants when it was kept hovering over them (p < 0.01). Notably, breathing status was misinterpreted as “no breathing” in 8 out of 27 (29.6%) participants lying in the supine position and 13 out of 36 (36.1%) participants lying in the prone position when the drone was kept hovering over them. The landing points seemed wider laterally when the participants were in the supine position than when they were in the prone position. Breathing status was more reliably determined when a small drone was landed on an individual’s body than when it hovered over them.

Background: When a rescuer walks alongside a stretcher and compresses the patient’s chest, the rescuer produces low-quality chest compressions. We hypothesized that a stretcher equipped with wing boards allows for better chest compressions than the conventional method. Methods: In this prospective, randomized, crossover study, we enrolled 45 medical workers and students. They performed hands-on chest compressions to a mannequin on a moving stretcher, while either walking (the walk method) or riding on wings attached to the stretcher (the wing method). The depths of the chest compressions were recorded. The participants’ vital signs were measured before and after the trials. Results: The average compression depth during the wing method (5.40 ± 0.50 cm) was greater than during the walk method (4.85 ± 0.80 cm; p < 0.01). The average compression rates during the two minutes were 215 ± 8 and 217 ± 5 compressions in the walk and wing methods, respectively (p = ns). Changes in blood pressure (14 ± 11 vs. 22 ± 14 mmHg), heart rate (32 ± 13 vs. 58 ± 20 bpm), and modified Borg scale (4 (interquartile range: 2–4) vs. 6 (5–7)) were significantly lower in the wing method cohort compared to the walking cohort (p < 0.01). The rescuer’s size and physique were positively correlated with the chest compression depth during the walk method; however, we found no significant correlation in the wing method. Conclusions: Chest compressions performed on the stretcher while moving using the wing method can produce high-quality chest compressions, especially for rescuers with a smaller size and physique.

Background: Conventional hands-on chest compression, in cardiopulmonary resuscitation, is often inadequate, especially when the rescuers are weak or have a small physique. Objectives: This study aimed to investigate the potential of leg-foot chest compression, with and without a footstool, during cardiopulmonary resuscitation. Methods and Results: We prospectively enrolled 21 medical workers competent in basic life support. They performed cardiopulmonary resuscitation on a manikin for 2 min using conventional hands-on compression (HO), leg-foot compression (LF), and leg-foot compression with a footstool (LF + FS). We analyzed the compression depths, changes in the rescuers’ vital signs, and the modified Borg scale scores after the trials. The compression depth did not differ between the cases using HO and LF. In the case of LF + FS, compression depths ⩾5 cm were more frequently observed (median, inter-quartile range: 93%, 81%–100%) than in HO (9%, 0%–57%, p < 0.01) and LF (28%, 11%–47%, p < 0.01). The increase in the heart rate or modified Borg scale scores, after the trials, did not differ between the HO and LF group; however, the values were the lowest in the case of LF + FS (49 ± 18 beats/min and 5 (4–7) in HO, 46 ± 18 and 6 (5–7) in LF, and 32 ± 11 and 2 (1–3) in LF + FS, respectively, p < 0.01). However, the increase in blood pressure, SpO₂, and respiratory rate were not different among each group. The increases in the heart rate and modified Borg scale scores negatively were correlated with the rescuers’ body size, in the case of HO and LF, but not LF + FS. Conclusion: LF can be used as an alternative to HO, when adequate HO is difficult. LF + FS could be used when rescuers are weak or have a small physique and when the victims are bigger than the rescuers.