Convolutional neural network-based vocal cord tumor classification technique for home-based self-prescreening purpose.

BACKGROUND: In this study, we proposed a deep learning technique that can simultaneously detect suspicious positions of benign vocal cord tumors in laparoscopic images and classify the types of tumors into cysts, granulomas, leukoplakia, nodules and polyps. This technique is useful for simplified home-based self-prescreening purposes to detect the generation of tumors around the vocal cord early in the benign stage. RESULTS: We implemented four convolutional neural network (CNN) models (two Mask R-CNNs, Yolo V4, and a single-shot detector) that were trained, validated and tested using 2183 laryngoscopic images. The experimental results demonstrated that among the four applied models, Yolo V4 showed the highest F1-score for all tumor types (0.7664, cyst; 0.9875, granuloma; 0.8214, leukoplakia; 0.8119, nodule; and 0.8271, polyp). The model with the lowest false-negative rate was different for each tumor type (Yolo V4 for cysts/granulomas and Mask R-CNN for leukoplakia/nodules/polyps). In addition, the embedded-operated Yolo V4 model showed an approximately equivalent F1-score (0.8529) to that of the computer-operated Yolo-4 model (0.8683). CONCLUSIONS: Based on these results, we conclude that the proposed deep-learning-based home screening techniques have the potential to aid in the early detection of tumors around the vocal cord and can improve the long-term survival of patients with vocal cord tumors.

Automated laryngeal mass detection algorithm for home-based self-screening test based on convolutional neural network.

BACKGROUND: Early detection of laryngeal masses without periodic visits to hospitals is essential for improving the possibility of full recovery and the long-term survival ratio after prompt treatment, as well as reducing the risk of clinical infection. RESULTS: We first propose a convolutional neural network model for automated laryngeal mass detection based on diagnostic images captured at hospitals. Thereafter, we propose a pilot system, composed of an embedded controller, a camera module, and an LCD display, that can be utilized for a home-based self-screening test. In terms of evaluating the model's performance, the experimental results indicated a final validation loss of 0.9152 and a F1-score of 0.8371 before post-processing. Additionally, the F1-score of the original computer algorithm with respect to 100 randomly selected color-printed test images was 0.8534 after post-processing while that of the embedded pilot system was 0.7672. CONCLUSIONS: The proposed technique is expected to increase the ratio of early detection of laryngeal masses without the risk of clinical infection spread, which could help improve convenience and ensure safety of individuals, patients, and medical staff.

A single-pieced, fully air-driven, cuff-inserted pseudo-blood pressure generator for on-site pre-screening test of non-invasive blood pressure monitor by nurses.

BACKGROUND: It is crucial to frequently inspect the proper operation of non-invasive electronic blood pressure monitors in various sites to prevent accidents from inaccurate blood pressure measurements, especially for large-scale hospitals. However, most conventional blood pressure monitor inspection devices are not suitable for such on-site investigation purpose. In this study, we propose a new single-pieced, fully air-driven pseudo blood pressure generator that is suitable for frequent on-site pre-screening tests of the blood pressure monitor by nurses. RESULTS: The proposed model comprises a rigid cylindrical body, two simulated brachial arteries, two air-pumps, an electronic controller, and a pressure sensor. Control algorithm based on polynomial curve fitting was implemented to generate various user-instructed systolic blood pressure and heart-rate conditions automatically. To evaluate the performance and clinical feasibility of the proposed model, various experiments were performed using ten commercial electronic blood pressure monitors. Experimental results demonstrated that the values of the Pearson coefficient between the reference pseudo-blood pressure waveforms and the actually generated pressure waveforms were 0.983, 0.983 and 0.997 at 60, 70 and 80 beats/min, respectively (p < 0.05). Besides, during the experiments using ten commercial blood pressure monitors, the maximum error in average systolic blood pressure was 2.9 mmHg, the maximum standard deviation in average systolic blood pressure was 3.5 mmHg, and the maximum percentage error in average pumping rate was 3.2%, respectively. CONCLUSIONS: We expect that the proposed model can give an easy and comprehensive way for frequent on-site investigations of the blood pressure monitors by nurses, and improve the safety of patients with abnormal blood pressure, especially in most large-scale hospitals.

Estimation of the variations in mechanical impedance between the actuator and the chest, and the power delivered to the chest during cardiopulmonary resuscitation using machine-embedded sensors.

BACKGROUND: To reduce the risk of patient damage and complications during the cardiopulmonary resuscitation (CPR) process in emergency situations, it is necessary to monitor the status of the patient and the quality of CPR while CPR processing without additional bio-signal measurement devices. In this study, an algorithm is proposed to estimate the mechanical impedance (MI) between an actuator of the CPR machine and the chest of the patient, and to estimate the power delivered to the chest of the patient during the CPR process. METHODS: Two sensors for force and depth measurement were embedded into a custom-made CPR machine and the algorithm for MI and power estimation was implemented. The performance of the algorithm was evaluated by comparing the results from the kinetic model, the conventional discrete Fourier transform (DFT), and the proposed method. RESULTS: The estimations of the proposed method showed similar increasing/decreasing trends with the calculations from the kinetic model. In addition, the proposed method showed statistically equivalent performance in the MI estimation, and at the same time, showed statistically superior performance in the power estimation compared with the calculations from the conventional DFT. Furthermore, the MI and power estimation could be performed almost in real-time during the CPR process without excessive hands-off periods, and the intensity of random noise contained in the input signals did not seriously affect the MI and power estimations of the proposed method. CONCLUSION: We expect that the proposed algorithm can reduce various CPR-related complications and improve patient safety.

An Eye-Blinking-Based Beamforming Control Protocol for Hearing Aid Users With Neurological Motor Disease or Limb Amputation.

For hearing-impaired individuals with neurological motor deficits or finger/arm amputation due to accident or disease, hearing aid adjustment using a conventional finger manipulation-based remote controller is unavailable, and a more dedicated, hands-free alternative is required. In this study, we propose an eye-blinking-based beamforming control scheme for hearing aid users. Three electroencephalogram signals measured around the ears were utilized to detect eye-blinking patterns based on a three-layer artificial neural network. The performance of the proposed control scheme was evaluated by both subjective experiments and objective index comparison tests in simulated situations. Experimental results from the subjective test demonstrated that without the pretraining phase, the accuracy and latency time were 68.57 ± 18.50% and 10.06 ± 0.94 s, respectively; in contrast, after the pretraining phase, both the accuracy and latency time were improved to 91.00 ± 4.69% and 8.60 ± 1.05 s, respectively. In index comparison tests, the proposed control scheme exhibited improvements in the signal-to-noise ratio (SNR) as well as the segmental SNR in all tested situations, as compared to a conventional forward-focusing beamforming algorithm. We believe that the proposed control scheme provides a novel, hands-free way in which to control the operation of hearing aids for hearing-impaired patients with additional motor deficits or amputation.

A Supporting Platform for Semi-Automatic Hyoid Bone Tracking and Parameter Extraction from Videofluoroscopic Images for the Diagnosis of Dysphagia Patients.

Conventional kinematic analysis of videofluoroscopic (VF) swallowing image, most popular for dysphagia diagnosis, requires time-consuming and repetitive manual extraction of diagnostic information from multiple images representing one swallowing period, which results in a heavy work load for clinicians and excessive hospital visits for patients to receive counseling and prescriptions. In this study, a software platform was developed that can assist in the VF diagnosis of dysphagia by automatically extracting a two-dimensional moving trajectory of the hyoid bone as well as 11 temporal and kinematic parameters. Fifty VF swallowing videos containing both non-mandible-overlapped and mandible-overlapped cases from eight patients with dysphagia of various etiologies and 19 videos from ten healthy controls were utilized for performance verification. Percent errors of hyoid bone tracking were 1.7 ± 2.1% for non-overlapped images and 4.2 ± 4.8% for overlapped images. Correlation coefficients between manually extracted and automatically extracted moving trajectories of the hyoid bone were 0.986 ± 0.017 (X-axis) and 0.992 ± 0.006 (Y-axis) for non-overlapped images, and 0.988 ± 0.009 (X-axis) and 0.991 ± 0.006 (Y-axis) for overlapped images. Based on the experimental results, we believe that the proposed platform has the potential to improve the satisfaction of both clinicians and patients with dysphagia.

A Diagonal-Steering-Based Binaural Beamforming Algorithm Incorporating a Diagonal Speech Localizer for Persons With Bilateral Hearing Impairment.

Previously suggested diagonal-steering algorithms for binaural hearing support devices have commonly assumed that the direction of the speech signal is known in advance, which is not always the case in many real circumstances. In this study, a new diagonal-steering-based binaural speech localization (BSL) algorithm is proposed, and the performances of the BSL algorithm and the binaural beamforming algorithm, which integrates the BSL and diagonal-steering algorithms, were evaluated using actual speech-in-noise signals in several simulated listening scenarios. Testing sounds were recorded in a KEMAR mannequin setup and two objective indices, improvements in signal-to-noise ratio (SNRi ) and segmental SNR (segSNRi ), were utilized for performance evaluation. Experimental results demonstrated that the accuracy of the BSL was in the 90-100% range when input SNR was -10 to +5 dB range. The average differences between the γ-adjusted and γ-fixed diagonal-steering algorithms (for -15 to +5 dB input SNR) in the talking in the restaurant scenario were 0.203-0.937 dB for SNRi and 0.052-0.437 dB for segSNRi , and in the listening while car driving scenario, the differences were 0.387-0.835 dB for SNRi and 0.259-1.175 dB for segSNRi . In addition, the average difference between the BSL-turned-on and the BSL-turned-off cases for the binaural beamforming algorithm in the listening while car driving scenario was 1.631-4.246 dB for SNRi and 0.574-2.784 dB for segSNRi . In all testing conditions, the γ-adjusted diagonal-steering and BSL algorithm improved the values of the indices more than the conventional algorithms. The binaural beamforming algorithm, which integrates the proposed BSL and diagonal-steering algorithm, is expected to improve the performance of the binaural hearing support devices in noisy situations.

A new asymmetric directional microphone algorithm with automatic mode-switching ability for binaural hearing support devices.

For hearing support devices, it is important to minimize the negative effect of ambient noises for speech recognition but also, at the same time, supply natural ambient sounds to the hearing-impaired person. However, conventional fixed bilateral asymmetric directional microphone (DM) algorithms cannot perform in such a way when the DM-mode device and a dominant noise (DN) source are placed on the same lateral hemisphere. In this study, a new binaural asymmetric DM algorithm that can overcome the defects of conventional algorithms is proposed. The proposed algorithm can estimate the position of a specific DN in the 90°-270° range and switch directional- and omnidirectional-mode devices automatically if the DM-mode device and the DN are placed in opposite lateral hemispheres. Computer simulation and KEMAR mannequin recording tests demonstrated that the performance of the conventional algorithm deteriorated when the DM-mode device and the DN were placed in the opposite hemisphere; in contrast, the performance of the proposed algorithm was consistently maintained regardless of directional variations in the DN. Based on these experimental results, the proposed algorithm may be able to improve speech quality and intelligibility for hearing-impaired persons who have similar degrees of hearing impairment in both ears.

An environment-adaptive management algorithm for hearing-support devices incorporating listening situation and noise type classifiers.

In order to provide more consistent sound intelligibility for the hearing-impaired person, regardless of environment, it is necessary to adjust the setting of the hearing-support (HS) device to accommodate various environmental circumstances. In this study, a fully automatic HS device management algorithm that can adapt to various environmental situations is proposed; it is composed of a listening-situation classifier, a noise-type classifier, an adaptive noise-reduction algorithm, and a management algorithm that can selectively turn on/off one or more of the three basic algorithms-beamforming, noise-reduction, and feedback cancellation-and can also adjust internal gains and parameters of the wide-dynamic-range compression (WDRC) and noise-reduction (NR) algorithms in accordance with variations in environmental situations. Experimental results demonstrated that the implemented algorithms can classify both listening situation and ambient noise type situations with high accuracies (92.8-96.4% and 90.9-99.4%, respectively), and the gains and parameters of the WDRC and NR algorithms were successfully adjusted according to variations in environmental situation. The average values of signal-to-noise ratio (SNR), frequency-weighted segmental SNR, Perceptual Evaluation of Speech Quality, and mean opinion test scores of 10 normal-hearing volunteers of the adaptive multiband spectral subtraction (MBSS) algorithm were improved by 1.74 dB, 2.11 dB, 0.49, and 0.68, respectively, compared to the conventional fixed-parameter MBSS algorithm. These results indicate that the proposed environment-adaptive management algorithm can be applied to HS devices to improve sound intelligibility for hearing-impaired individuals in various acoustic environments.

Effect of counter-pulsation control of a pulsatile left ventricular assist device on working load variations of the native heart.

BACKGROUND: When using a pulsatile left ventricular assist device (LVAD), it is important to reduce the cardiac load variations of the native heart because severe cardiac load variations can induce ventricular arrhythmia. In this study, we investigated the effect of counter-pulsation control of the LVAD on the reduction of cardiac load variation. METHODS: A ventricular electrocardiogram-based counter-pulsation control algorithm for a LVAD was implemented, and the effects of counter-pulsation control of the LVAD on the reduction of the working load variations of the left ventricle were determined in three animal experiments. RESULTS: Deviations of the working load of the left ventricle were reduced by 51.3%, 67.9%, and 71.5% in each case, and the beat-to-beat variation rates in the working load were reduced by 84.8%, 82.7%, and 88.2% in each ease after counter-pulsation control. There were 3 to 12 premature ventricle contractions (PVCs) before counter-pulsation control, but no PVCs were observed during counter-pulsation control. CONCLUSIONS: Counter-pulsation control of the pulsatile LVAD can reduce severe cardiac load variations, but the average working load is not markedly affected by application of counter-pulsation control because it is also influenced by temporary cardiac outflow variations. We believe that counter-pulsation control of the LVAD can improve the long-term safety of heart failure patients equipped with LVADs.

Enhanced beam-steering-based diagonal beamforming algorithm for binaural hearing support devices.

In order to improve speech intelligibility for hearing-impaired people in various listening situations, it is necessary to diversify the possible focusing directions of a beamformer. In a previous report, the concept of binaural beam-steering that can focus a beamformer in diagonal directions was applied to a binaural hearing aid; however, in the previously proposed protocol, the effective frequency range for consistent diagonal beam-steering was limited to the 200-750 Hz range, which is far narrower than that of normal speech signals (200-4000 Hz). In this study, we proposed a modified binaural diagonal beam-steering technique that can reduce the focusing-direction deviations at high input frequencies up to 4000 Hz by introducing a new correction factor to the original protocol that can reduce the differences in gradient between the signal and the noise components at frequencies up to 4000 Hz. In simulation tests, the focusing effect of the proposed algorithm was more consistent than conventional algorithms. The deviations between the target and the focusing directions were reduced 27% in the left device and 6% in the right device with 45° steering at a 4000 Hz input signal, and were reduced 3% in the left device and 25% in the right device with 135° steering. On the basis of the experimental results, we believe that the proposed algorithm has the potential to help hearing-impaired people in various listening situations.

In vitro comparative study of wall portbased high-frequency chest wall oscillation device and internal air-pulse generator device.

Owing to environmental and disease issues, the use of high-frequency chest wall oscillation (HFCWO) devices in hospitals is consistently increasing. This study proposes a cost-effective actuator-less HFCWO device that utilizes an external wall port utility in hospitals to generate the positive and negative pneumatic pressures required for HFCWO treatment instead of an embedded mechanical actuator. The manufactured prototype with the no-amplification (NO-AMP) setting contained an electric pressure regulator to enable intensity level adjustment and two solenoid valves to enable vibration frequency adjustment, whereas the prototype with the pre-amplification (PRE-AMP) setting contained an additional air reservoir and an air-pressure booster. The prototype device was tuned to output average local maximum values in the pressure waveform similar to a commercial VEST-205 device at an 8-12 Hz frequency and 2-4 pressure intensity levels. In vitro comparative experiments demonstrated that the prototype device showed similar local maximum pressures to those of the VEST-205 (mean absolute pressure difference, <3 mmH2O); in contrast, the proposed device showed significantly higher local minimum pressures than those of the VEST-205 (mean absolute pressure difference, >8 mmH2O). Additionally, the driving sound of the proposed device was 17.0-17.8 dB higher than that of VEST-205. We conclude that the proposed device has the potential to substitute for conventional HFCWO devices under the limited but most frequently used operating conditions, although more detailed modifications are necessary in future studies to improve its performance and clinical usability.

New computerized indices for quantitative evaluation of depression and asymmetry in patients with chest wall deformities.

An evaluation index that can quantitatively assess the severity of chest wall deformities is essential to prepare and assess corrective surgical operations for patients with these deformities, including funnel chest patients. In previous studies, our group proposed several automatically calculated indices that represent the severity of depression and asymmetry in the chest wall. These indices showed sufficient performance in most cases of deformities, including those involving asymmetric and symmetric depression; however, their linearity declined when assessing complex deformities. The purpose of this study is to propose two automated indices that provide linear evaluation output for all types of chest wall deformities, including complex deformities, and to evaluate their performance and clinical feasibility. Six reference chest wall boundary curves were obtained from 60 computed tomography (CT) images of a normal chest. Next, an active contour model-based image processing technique was used to extract boundary curves from images of patients with real chest wall deformities. Third, the required parameters were extracted from the boundary curves and the targeted indices were calculated. Finally, the performance of the proposed indices was evaluated using 33 synthetic images and 60 real chest CT images of patients with chest wall deformities. The newly proposed indices can be automatically calculated from the original CT images and showed sufficient performance for all types of chest wall deformities. We believe that the newly proposed indices can facilitate pre- and postoperative evaluation of chest wall deformities in clinical practice.

Deep learning-based monitoring technique for real-time intravenous medication bag status.

Accidents related to the administration of intravenous (IV) medication, such as drug overdose/underdose, drug/patient mis-identification, and delayed bag exchange, occur consistently in clinical fields. Several previous studies have suggested various contact-sensing and image-processing methodologies; however, most of them can increase the workload of nursing staffs during the long-term, continuous monitoring. In this study, we proposed a smart IV pole that can monitor the infusion status of up to four IV medications (patient/drug identification, and liquid residue) with various sizes and hanging positions to reduce IV-related accidents and improve patient safety with the least additional workload; the system consists of 12 cameras, one code scanner, and four controllers. Two types of deep learning models for automated camera selection (CNN-1) and liquid residue monitoring (CNN-2), and three drug residue estimation equations were implemented. The experimental results demonstrated that the accuracy of identification code-checking (60 tests) was 100%. The classification accuracy and the mean inference time of CNN-1 (1200 tests) were 100% and 140 ms. The mean average precision and the mean inference time of CNN-2 (300 tests) were 0.94 and 144 ms. The average error rates between the alarm setting (20, 30, and 40 mL) and the actual drug residue when the alarm first generated were 4.00%, 7.33%, and 4.50% for a 1,000 mL bag; 6.00%, 4.67%, and 2.50% for a 500 mL bag; and 3.00%, 6.00%, and 3.50% for a 100 mL bag, respectively. Our results suggest that the implemented AI-based prototype IV pole is a potential tool for reducing IV-related accidents and improving in-hospital patient safety. Supplementary Information: The online version contains supplementary material available at 10.1007/s13534-023-00292-w.

A novel endoscopic fluorescent clip for the localization of gastrointestinal tumors.

BACKGROUND: Accurate tumor localization is essential for gastrointestinal surgery, especially in cases of early cancer. This study was designed to develop a novel fluorescent clip for rapid and exact visualization of tumor sites. METHODS: A transparent polymer matrix containing highly bright fluorochromes was coated on the front end of endoscopic clips. The fluorescent clips were placed on the mucosal surface of a porcine colon and stomach, and the operator then attempted to identify the fluorescent clips from the outer serosal side of the colon and stomach. A 532-nm diode laser and filter glass were used for visualizing the fluorescence signals through the colonic tissue. A 650-nm diode laser and a digital charge-coupled device (CCD) camera equipped with a bandpass emission filter were used for the imaging of the fluorescent clips through the thick stomach tissue. RESULTS: When a green light from a 532-nm diode laser (power density=0.35 mW/cm2) was applied on the serosal surface of the porcine colon, we could identify all clips that had been placed endoscopically on the mucosal surface of the inner colonic wall. By using the light from a 650-nm diode laser (power density=0.7 mW/cm2), we identified all fluorescent clips through the stomach wall in real time. Similar results were also obtained with the filtered xenon lamp. CONCLUSION: An endoscopic fluorescent clip can be useful for the rapid and exact localization of tumors, and this technique can also be useful during laparoscopic surgery.

Single-dose volume regulation algorithm for a gas-compensated intrathecal infusion pump.

The internal pressures of medication reservoirs of gas-compensated intrathecal medication infusion pumps decrease when medication is discharged, and these discharge-induced pressure drops can decrease the volume of medication discharged. To prevent these reductions, the volumes discharged must be adjusted to maintain the required dosage levels. In this study, the authors developed an automatic control algorithm for an intrathecal infusion pump developed by the Korean National Cancer Center that regulates single-dose volumes. The proposed algorithm estimates the amount of medication remaining and adjusts control parameters automatically to maintain single-dose volumes at predetermined levels. Experimental results demonstrated that the proposed algorithm can regulate mean single-dose volumes with a variation of <3% and estimate the remaining medication volume with an accuracy of >98%.

A durability study of a paracorporeal pulsatile electro-mechanical pneumatic biventricular assist device.

In 2002, the paracorporeal pulsatile electro-mechanical pneumatic ventricular assist device (VAD) began to be developed by the Korea Artificial Organ Center at Korea University under a Health & Medical Technology Research and Development program which finished in 2008. In vitro durability testing was conducted on the paracorporeal pulsatile pneumatic VAD to determine device durability and to evaluate device failures. The 1- and 2-year reliability of the paracorporeal pulsatile pneumatic VAD was shown to be 91.2% and 54.9%, respectively, with an 80% confidence level. Failure modes were analyzed using fault tree analysis, with customized software continuously acquiring data during the test period. After this period, 21 in vivo animal tests were done, with 14 cases of left atrium to left ventricle (LV) inflow cannulation (36Fr)/outflow grafting to descending aorta, and seven cases of apex cannulation of LV to descending aorta (12 mm). The longest postoperative day (182 days) in Korea was recently recorded in in vivo animal testing (bovine, 90 kg, male, 3.5-4.0 L/min flow rate, and 55 bpm).

Multi-modal infusion pump real-time monitoring technique for improvement in safety of intravenous-administration patients.

Intravenous (IV) medication administration processes have been considered as high-risk steps, because accidents during IV administration can lead to serious adverse effects, which can deteriorate the therapeutic effect or threaten the patient's life. In this study, we propose a multi-modal infusion pump (IP) monitoring technique, which can detect mismatches between the IP setting and actual infusion state and between the IP setting and doctor's prescription in real time using a thin membrane potentiometer and convolutional-neural-network-based deep learning technique. During performance evaluation, the percentage errors between the reference infusion rate (IR) and average estimated IR were in the range of 0.50-2.55%, while those between the average actual IR and average estimated IR were in the range of 0.22-2.90%. In addition, the training, validation, and test accuracies of the implemented deep learning model after training were 98.3%, 97.7%, and 98.5%, respectively. The training and validation losses were 0.33 and 0.36, respectively. According to these experimental results, the proposed technique could provide improved protection functions to IV-administration patients.

Convolutional neural network-based ambient light-independent panel digit surveillance technique for infusion pumps.

For effective patient therapy and improved patient safety, it is critical to administer medication accurately in accordance with doctor's prescription. However, accidents owing to the erroneous programing of infusion pumps caused by users have been consistently reported in several documents. In this study, the authors propose a novel surveillance technique for infusion pumps to continuously monitor the variations in panel digits using a convolutional neural network model, and evaluate the performance of the implemented technique. During the experimental evaluation, 1st-step ROIs and 2nd-step ROIs were successfully extracted from the frame images regardless of the ambient lighting conditions. The final accuracies of the implemented CNN model are 99.9% for both the training (172,800 images) and validation (1080 images) dataset while the final losses for the training and validation datasets are 0.48 and 0.45 after 13th epoch, respectively. In the 24-h continuous monitoring test, the accuracy of the model for volume recognition considering all the 1440 measurements (960 for day-lighting and 480 for night-lighting) is 95.5%, whereas in day-lighting and night-lighting modes the accuracies of the model are 98.2% and 90.0%, respectively. Based on these experimental results, the proposed surveillance technique incorporating infusion pumps is expected to improve the safety of patients who need long-term treatments via infusion pumps, reducing the burden on the nurses and hospitals.

Optimal moving angle of pusher plate in occlusive-type pulsatile blood pump.

Since the occlusive-type pulsatile extracorporeal blood pump (Twin-Pulse Life Support System; Seoul National University, Seoul, Korea) received the CE mark of the European Directives and Korea Food and Drug Administration approval (2004) for short-term applications as an extracorporeal life support system, the pump system has been tested for hemolysis. This pump system was recently upgraded with an ameliorated pusher plate to reduce hemolysis. In this study, numerical analysis and in vitro tests were performed to determine the optimal conditions for increasing the durability of the blood sac and pump output. During the simulation, the minimum sliding interface force (SIF) for the angle of the pusher plate movement (PPM) was calculated (40-70 degrees ). In the in vitro durability test, the angle of the PPM was increased gradually from 40 to 70 degrees in 10 degrees increments, and the mean time to failure (MTTF) of the blood sac was calculated. Fifteen tests were conducted for each case: 40, 50, 60, and 70 degrees (n = 15 each). The MTTF of the blood sac was defined as the time when a crack of the blood sac occurred. The longer lifetime of the blood sac at 60 degrees of the PPM (297.0 h) than that at 50 degrees (197.6 h) was attributed to the lower SIF value (-0.13, normalized value) at 60 degrees of the PPM.