Retrograde cardioplegia is now an alternative or adjunctive method used worldwide as a cardiac protection during open heart surgery. However, its use involves some limitation. We operated on a patient suffering from aortic stenosis associated with PLSVC. During the operation on this patient for aortic valve replacement, retrograde infusion of cardioplegic solution could not be performed because the coronary sinus was excessively dilated and prevented the balloon from occluding it. Other anomalous lesion of the coronary sinus make the retrograde infusion of the cardioplegic solution difficult and these must always be kept in mind when cardioplegia is infused from the coronary sinus.

A 64-year-old man with a history of anterior blunt trauma 10 years previously was admitted to our hospital complaining of general fatigue. A plain chest roentgenogram showed pericardial calcification. Computed tomography and echocardiography showed the mass to be a calcified capsule in the anterior mediastinum compressing the right side of the heart. He underwent an operation through a median sternotomy. The mass was an organized hematoma encapsulated by a calcified fibrous and serous layer of the pericardium. The hematoma was resected together with the calcified pericardium under cardiopulmonary bypass. His postoperative course was uneventful. He had no history of hemopericardium but had experienced blunt chest trauma that seemed to have induced the subsequent localized constrictive pericarditis.

Perivalvular leakage (PVL) is one of the serious complications of mitral valve replacement. Between 1991 and 2006, 9 patients with mitral PVL underwent reoperation. All of them had severe hemolytic anemia before surgery. The serum lactate dehydrogenase (LDH) level decreased from 2,366±780 IU/l to 599±426 IU/l after surgery. The site of PVL was accurately defined in 7 patients by echocardiography. PVL occurred around the posterior annulus in 3 patients, anterior annulus in 2, anterolateral commissure in 1, and posteromedial commissure in 1. The most frequent cause of PVL was annular calcification in 5 patients. Infection was only noted in 1 patient. In 4 patients, the prosthesis was replaced, while the leak was repaired in 5 patients. There was one operative death, due to multiple organ failure, and 4 late deaths. The cause of late death was cerebral infarction in 1 patient, subarachnoid hemorrhage in 1, sudden death in 1, and congestive heart failure (due to persistent PVL) in 1. Reoperation for PVL due to extensive annular calcification is associated with a high mortality rate and high recurrence rate, making this procedure both challenging and frustrating for surgeons.

Objectives: This study evaluated the efficacy of integrating a retrieval-augmented generation (RAG) model and a large language model (LLM) to improve the accuracy of drug name mapping across international vocabularies. Methods: Drug ingredient names were translated into English using the Japanese Accepted Names for Pharmaceuticals. Drug concepts were extracted from the standard vocabulary of OHDSI, and the accuracy of mappings between translated terms and RxNorm was assessed by vector similarity, using the BioBERT-generated embedded vectors as the baseline. Subsequently, we developed LLMs with RAG that distinguished the final candidates from the baseline. We assessed the efficacy of the LLM with RAG in candidate selection by comparing it with conventional methods based on vector similarity. Results: The evaluation metrics demonstrated the superior performance of the combined LLM + RAG over traditional vector similarity methods. Notably, the hit rates of the Mixtral 8x7b and GPT-3.5 models exceeded 90%, significantly outperforming the baseline rate of 64% across stratified groups of PO drugs, injections, and all interventions. Furthermore, the r-precision metric, which measures the alignment between model judgment and human evaluation, revealed a notable improvement in LLM performance, ranging from 41% to 50% compared to the baseline of 23%. Conclusions Integrating an RAG and an LLM outperformed conventional string comparison and embedding vector similarity techniques, offering a more refined approach to global drug information mapping.

Objectives: This study evaluated the efficacy of integrating a retrieval-augmented generation (RAG) model and a large language model (LLM) to improve the accuracy of drug name mapping across international vocabularies. Methods: Drug ingredient names were translated into English using the Japanese Accepted Names for Pharmaceuticals. Drug concepts were extracted from the standard vocabulary of OHDSI, and the accuracy of mappings between translated terms and RxNorm was assessed by vector similarity, using the BioBERT-generated embedded vectors as the baseline. Subsequently, we developed LLMs with RAG that distinguished the final candidates from the baseline. We assessed the efficacy of the LLM with RAG in candidate selection by comparing it with conventional methods based on vector similarity. Results: The evaluation metrics demonstrated the superior performance of the combined LLM + RAG over traditional vector similarity methods. Notably, the hit rates of the Mixtral 8x7b and GPT-3.5 models exceeded 90%, significantly outperforming the baseline rate of 64% across stratified groups of PO drugs, injections, and all interventions. Furthermore, the r-precision metric, which measures the alignment between model judgment and human evaluation, revealed a notable improvement in LLM performance, ranging from 41% to 50% compared to the baseline of 23%. Conclusions Integrating an RAG and an LLM outperformed conventional string comparison and embedding vector similarity techniques, offering a more refined approach to global drug information mapping.

Objectives: This study evaluated the efficacy of integrating a retrieval-augmented generation (RAG) model and a large language model (LLM) to improve the accuracy of drug name mapping across international vocabularies. Methods: Drug ingredient names were translated into English using the Japanese Accepted Names for Pharmaceuticals. Drug concepts were extracted from the standard vocabulary of OHDSI, and the accuracy of mappings between translated terms and RxNorm was assessed by vector similarity, using the BioBERT-generated embedded vectors as the baseline. Subsequently, we developed LLMs with RAG that distinguished the final candidates from the baseline. We assessed the efficacy of the LLM with RAG in candidate selection by comparing it with conventional methods based on vector similarity. Results: The evaluation metrics demonstrated the superior performance of the combined LLM + RAG over traditional vector similarity methods. Notably, the hit rates of the Mixtral 8x7b and GPT-3.5 models exceeded 90%, significantly outperforming the baseline rate of 64% across stratified groups of PO drugs, injections, and all interventions. Furthermore, the r-precision metric, which measures the alignment between model judgment and human evaluation, revealed a notable improvement in LLM performance, ranging from 41% to 50% compared to the baseline of 23%. Conclusions Integrating an RAG and an LLM outperformed conventional string comparison and embedding vector similarity techniques, offering a more refined approach to global drug information mapping.

Objectives: This study evaluated the efficacy of integrating a retrieval-augmented generation (RAG) model and a large language model (LLM) to improve the accuracy of drug name mapping across international vocabularies. Methods: Drug ingredient names were translated into English using the Japanese Accepted Names for Pharmaceuticals. Drug concepts were extracted from the standard vocabulary of OHDSI, and the accuracy of mappings between translated terms and RxNorm was assessed by vector similarity, using the BioBERT-generated embedded vectors as the baseline. Subsequently, we developed LLMs with RAG that distinguished the final candidates from the baseline. We assessed the efficacy of the LLM with RAG in candidate selection by comparing it with conventional methods based on vector similarity. Results: The evaluation metrics demonstrated the superior performance of the combined LLM + RAG over traditional vector similarity methods. Notably, the hit rates of the Mixtral 8x7b and GPT-3.5 models exceeded 90%, significantly outperforming the baseline rate of 64% across stratified groups of PO drugs, injections, and all interventions. Furthermore, the r-precision metric, which measures the alignment between model judgment and human evaluation, revealed a notable improvement in LLM performance, ranging from 41% to 50% compared to the baseline of 23%. Conclusions Integrating an RAG and an LLM outperformed conventional string comparison and embedding vector similarity techniques, offering a more refined approach to global drug information mapping.

Methods: This study included 229 patients (107 men and 122 women; mean age, 68.9 years) who received one or two levels of MISTLIF, and the patient’s age, gender, disease, paralysis, preoperative physical functions, duration of symptom(s), and surgery-associated factors (waiting for surgery, number of surgical levels, surgical time, and intraoperative blood loss) were studied. Radiographic characteristics and clinical outcomes such as Oswestry Disability Index (ODI) scores and Visual Analog Scale (VAS; 0–100) ODI scores for low back pain, leg pain, and numbness were studied. One year following surgery, patient satisfaction (defined as satisfaction for surgery and for present condition; 0–100) was assessed using VAS and its relationships with investigation factors were examined. Results: The mean VAS scores of satisfaction for surgery and for present condition were 88.6 and 84.2, respectively. The results of multiple regression analysis showed that preoperative adverse factors of satisfaction for surgery were being elderly (β =-0.17, p =0.023), high preoperative low back pain VAS scores (β =-0.15, p =0.020), and postoperative adverse factors were high postoperative ODI scores (β =-0.43, p <0.001). In addition, the preoperative adverse factor of satisfaction for present condition was high preoperative low back pain VAS scores (β =-0.21, p =0.002), and postoperative adverse factors were high postoperative ODI scores (β =-0.45, p <0.001) and high postoperative low back pain VAS scores (β =-0.26, p =0.001). Conclusions According to this study, significant preoperative low back pain and high postoperative ODI score after surgery are linked to patient unhappiness.

Pathogens handled in a Biosafety Level 3 (BSL-3) containment laboratory pose significant risks to laboratory staff and the environment. It is therefore necessary to develop competency and proficiency among laboratory workers and to promote appropriate behavior and practices that enhance safety through biosafety training. Following the installation of our BSL-3 laboratory at the Center for Microbiology Research-Kenya Medical Research Institute in 2006, a biosafety training program was developed to provide training on BSL-3 safety practices and procedures. The training program was developed based on World Health Organization specifications, with adjustments to fit our research activities and biosafety needs. The program is composed of three phases, namely initial assessment, a training phase including theory and a practicum, and a final assessment. This article reports the content of our training program.

Pathogens handled in a Biosafety Level 3 (BSL-3) containment laboratory pose significant risks to laboratory staff and the environment. It is therefore necessary to develop competency and proficiency among laboratory workers, and promote behaviors and practices that enhance safety through biosafety training. Following installation of our BSL-3 laboratory at the Center for Microbiology Research-Kenya Medical Research Institute, in 2006, a biosafety training program was developed to provide training on BSL-3 safety practices and procedures. The training program was developed based on the World Health Organization specifications, with adjustments to fit our research activities and biosafety needs. The program is composed of three phases namely; initial assessment, a training phase that includes theory and practicum, and final assessment. This article reports the content of our training program.