Yoji Kishi

BACKGROUND Background: Large language models (LLMs) may support clinical documentation; however, comparative evidence on Japanese subjective, objective, assessment, and plan (SOAP) note generation from transcripts of actual outpatient encounters remains limited. OBJECTIVE Objective: We aimed to compare three LLMs, across three generation runs, for Japanese SOAP note generation from automatically transcribed outpatient encounters and to assess the agreement between physician ratings and a language-model judge. METHODS Methods: This single-center retrospective evaluation used 154 deidentified automatic speech recognition (ASR) transcripts from three outpatient departments in Japan. Qwen2.5-14B-Instruct, Gemini 2.5 Pro, and Claude Opus 4.6 each generated three notes per transcript in three independent runs under common instructions (1386 notes), all of which were evaluated. Two physicians created reference notes from the same ASR text; the original audio was unavailable. Three other physicians, blinded to model identity, rated all notes on nine attributes of an adapted Provider Documentation Summarization Quality Instrument and screened them for four categories of severe documentation error. Models were compared using mixed-effects models. GPT-4o, used as an LLM-as-a-judge, scored eight attributes with generic and, in a second run, the physicians' anchors; agreement with mean physician ratings was assessed using absolute-agreement intraclass correlation coefficients (ICCs) and Bland-Altman analysis. Analyses were exploratory and unadjusted for multiplicity. RESULTS Results: Mean nine-item scores were 4.02 (SD 0.38) for Qwen, 4.26 (SD 0.33) for Gemini, and 4.38 (SD 0.32) for Claude (omnibus P<.001); the reliability of the three-physician mean was ICC(2,3)=0.84 (95% CI 0.81-0.87). At least one severe error was identified in 85/462 (18.4%; 95% CI 15.1%-22.2%), 57/462 (12.3%; 95% CI 9.6%-15.7%), and 35/462 (7.6%; 95% CI 5.5%-10.4%) notes, respectively (omnibus P<.001); the 95% CIs of all three model-based pairwise odds ratios excluded 1, and critical omission was the most frequent category. The ordering of the models was the same in each run, and 33.8%, 23.4%, and 14.9% of transcripts had a severe error in at least one run. Severe errors were present in 18.6%, 11.1%, and 7.4% of notes rated clinically useful. The judge-physician ICC for the eight-item score was 0.68 (95% CI 0.61-0.74) with generic anchors and 0.71 (95% CI 0.65-0.76) with the physicians' anchors. Of the 177 notes with a severe error, 85 (48.0%) were attributed to features of speech-recognized dialogue, such as speaker-role misattribution, misresolution of omitted subjects in Japanese, and suspected speech recognition errors. CONCLUSIONS Conclusions: The models differed in physician-rated quality and in the proportion of notes with severe documentation errors, and severe errors occurred with every model in every run, including in notes rated useful. These findings support prospective evaluation of physician-reviewed drafting but do not establish the safety of complete audio-to-note pipelines or workflow benefits. Language-model scoring agreed moderately with physician ratings and may complement, but not replace, physician assessment.

Tumor size plays a central role in the staging, resectability assessment, and response evaluation of pancreatic ductal adenocarcinoma (PDAC). However, radiologic tumor size does not necessarily correspond to pathologic tumor size, and this discrepancy has become more clinically relevant in the preoperative therapy era. PDAC is characterized by infiltrative growth, abundant desmoplastic stroma, and ill-defined tumor borders, all of which complicate accurate size assessment on imaging. After preoperative therapy, fibrosis, necrosis, and scattered residual tumor cells further obscure the relationship between radiologic abnormalities and viable tumor burden. This review summarizes the mechanisms underlying radiologic–pathologic discrepancies in tumor size assessment of PDAC, with a focus on computed tomography, magnetic resonance imaging, ultrasound, and pathologic evaluation after preoperative therapy. We also discuss the prognostic implications of imaging-based and pathological tumor sizes and highlight the limitations of size-based response assessment. Finally, we review emerging approaches beyond conventional size metrics, including functional imaging, radiomics, and pathological standardization. Recognizing the biological and treatment-related basis of tumor size discrepancy is essential for improving tumor assessment and refining staging strategies in contemporary PDAC management.

Background With the widespread availability of whole-slide imaging, many studies have utilized digital images of hematoxylin and eosin (H&E)-stained breast cancer tissues and applied convolutional neural networks (CNNs) for pathological diagnosis. However, CNN-based diagnosis is largely a black box and may be limited in quantitative morphological research. In this study, we developed a simple algorithm for morphometric analysis of three nuclear atypia features on H&E-stained whole-slide images to predict nuclear grade, hormone receptor status, and Ki-67 levels in breast cancer. Materials and Methods Using 43,183 H&E-stained nuclear images larger than 20 µm² from 131 invasive ductal breast carcinomas, we calculated the following features of nuclear atypia using a computer vision algorithm: anisonucleosis (variation in nuclear size), inhomogeneous chromatin density, and the average size of prominent nucleoli. Anisonucleosis was quantified as the percentage of nuclei larger than 47 μm². Inhomogeneous chromatin was defined as the percentage of blue-saturated structures with 0.92-fold luminance or less than the average nuclear luminance. Prominent nucleoli were identified based on blue-saturated structures with 0.87-fold luminance or less, circularity greater than 0.65, and size greater than 1.15 μm². Using these values of nuclear atypia features, the thresholds that were the most associated with grade and biomarkers were calculated using receiver operating characteristic curves by Youden index. Results The morphometric algorithm using these thresholds predicted nuclear grade (grade 1 and 3), Ki-67 index ≧ 20%, and hormone receptor negative status with sensitivities of 52.1 to 100% and specificities from 34.3 to 85.3%. Two multivariable logistic regression models combining these three thresholds predicted nuclear grade, Ki-67, and hormone receptor negative with much better accuracy, sensitivities ranging from 57.1 to 91.3%, specificities from 50.9 to 82.2%, and area under curve of 0.70–0.82. The algorithm was applied to an independent set of 42 tumors. Conclusion The present morphological algorithm of nuclear atypia might provide new insights into the computational grading of invasive breast cancer.

We report a rare case of giant hepatocellular carcinoma (HCC) with direct invasion to the colon and synchronous multiple lung metastases that achieved complete remission with surgical hepatectomy followed by chemotherapy with atezolizumab and bevacizumab. An 80-year-old man presented with abdominal distension and appetite loss. Imaging studies demonstrated a large HCC protruding from the liver with compression and invasion of the ascending colon, accompanied by multiple bilateral pulmonary nodules. To relieve symptoms and reduce the risk of tumor rupture, upfront surgical resection consisting of right hepatectomy and right hemicolectomy was performed. Histopathological examination revealed moderately to poorly differentiated HCC with direct invasion into the ascending colon and microscopic portal vein invasion. Postoperatively, tumor markers initially decreased; however, subsequent elevation led to initiation of atezolizumab plus bevacizumab for treatment of pulmonary metastases. Tumor markers subsequently normalized, and follow-up computed tomography demonstrated disappearance of all pulmonary metastases, with only scar-like lesions remaining. Radiologic complete remission was achieved 14 months after hepatectomy. The patient remains alive without recurrence 54 months after surgery. Direct invasion of HCC into adjacent organs is extremely rare and is generally associated with poor prognosis. The role of systemic chemotherapy after reduction surgery for advanced HCC remains controversial. Although this is a single case experience and caution is needed when generalizing, this case suggested that combined treatment with initial resection of the primary tumor followed by immunotherapy for remnant lesions could be feasible in selected patients and would potentially contribute to symptom relief and control of metastatic lesions.

Pyloromyotomy is a technique for relieving infantile hypertrophic pyloric stenosis and is performed almost safely without severe complications during long-term outcomes. To the best of our knowledge, this is the first reported case of pancreatic cancer arising from the pancreatic parenchyma embedded within the pyloric ring after neonatal pyloromyotomy. A 39-year-old woman with a history of infantile hypertrophic pyloric stenosis underwent a pyloromyotomy during the neonatal period and presented with progressive nausea and vomiting. Imaging studies suggested the presence of ectopic pancreatic or gastric cancer invading the pancreas. After laparotomy, tumor invasion of the pancreatic head was identified; therefore, pancreaticoduodenectomy was performed. Pathological examination confirmed pancreatic cancer arising from the pancreatic parenchyma embedded within the duodenal and pyloric muscle layers. The patient completed 6 months of adjuvant chemotherapy with oral S-1. P eritoneal dissemination was observed 23 months postoperatively. Abnormal embedding of the pancreatic head in the walls of the duodenum and pyloric ring was considered to be related to neonatal pyloromyotomy. This alteration requires careful preoperative evaluation and flexible surgical planning to achieve curative resection of atypical presentations of gastrointestinal malignancies. The cancer in the present case may be associated with long-term anatomical alterations following neonatal pyloromyotomy, although a direct causal relationship cannot be established.

We present a novel triple-catheter adjunctive technique through a 6-Fr distal access catheter (DAC) for coil embolisation of an 18 mm wide-necked splenic aneurysm located at the bifurcation of the superior and inferior branches in a 63-year-old female. Despite the significantly tortuous splenic artery, the DAC was successfully navigated near the aneurysm, providing catheter stability. One 2.8-Fr balloon catheter and 2 2.2-Fr microcatheters were deployed simultaneously through the DAC to perform the balloon-assisted double-catheter technique. Dense packing was achieved with a volume embolisation ratio of 39.2%, while preserving the superior branch flow. Post-embolisation angiography revealed complete aneurysmal occlusion without splenic infarction. This technique presents a viable solution for complex visceral aneurysms in tortuous anatomies, enabling dense packing while preserving the parent artery, without requiring multiple access sites or larger guiding catheters.

A 78‐year‐old woman presented with sudden‐onset lower extremity edema, abdominal pain, and loss of appetite that developed the day after she fell off her bicycle. Computed tomography showed a simple hepatic cyst, 17 cm in diameter, compressing the inferior vena cava (IVC). Her symptoms resolved following laparoscopic deroofing. Notably, the cyst had remained stable at 12 cm 2 and 4 years prior, without any evidence of IVC compression. Although many guidelines for asymptomatic simple hepatic cysts published to date do not recommend treatment or follow‐up regardless of size, our experience suggests the importance of monitoring simple hepatic cysts larger than 12 cm.

Carbohydrate restriction (ketogenic diet) is a cancer treatment that reduces energy production by oxidative phosphorylation in mitochondria of cancer cells and increases it through anaerobic glycolysis in cytoplasm. We report a patient in whom progression of pancreatic cancer recurrence was suppressed solely by a ketogenic diet for 9-month post-surgery. A 60-year-old female with a diagnosis of pancreatic cancer underwent pancreatoduodenectomy after 3 cycles of chemotherapy with gemcitabine plus nab-paclitaxel. Multiple pulmonary metastases were observed 22 months after the surgery. We administered gemcitabine plus nab-paclitaxel for 1 year. As the partial response continued for 1 year, we performed radiotherapy for the remnant pulmonary metastases followed by administration of S-1 for 6 months. Ten months after radiotherapy, CT showed exacerbation of the pulmonary metastases. As treatment, she requested severe carbohydrate restriction. After 9 months of the ketogenic diet, CT revealed stable disease. A ketogenic diet may have the therapeutic effect of suppressing tumor progression if strictly applied.

Purpose This prospective study aimed to compare the diagnostic accuracy of hepatic lesions between contrast-enhanced computed tomography (CE-CT) and gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid-enhanced magnetic resonance imaging (Gd-EOB-MRI). We also examined the final diagnosis of equivocal lesions. Methods In each patient undergoing hepatectomy for liver tumors, 2 radiologists (readers A and B) assessed CE-CT and Gd-EOB-MRI. Hepatic lesions were scored as 1, 2, 3, 4, or 5, which corresponded to categorization as benign, probably benign, equivocal, probably malignant, or malignant, respectively. Those scored as 1/2 and 4/5 were defined as benign and malignant, respectively. The lesions left unresected were confirmed as benign when they remained unchanged for ≥ 1 year after hepatectomy. The false-positive rate was defined as the number of lesions finally confirmed as benign among those assessed as malignant preoperatively. Results Among 105 enrolled patients (colorectal liver metastases, 72; hepatocellular carcinoma, 29; other, 4), 886 lesions were recognized on either CT or MRI preoperatively; another 26 lesions were identified only on intraoperative ultrasonography or pathologic examination. The sensitivity of malignant lesion detection was significantly higher for Gd-EOB-MRI than CE-CT for both reader A (90.3% vs. 68.3%; P < 0.001) and reader B (88.1% vs. 70.2%; P < 0.001). The false-positive rate with CE-CT and Gd-EOB-MRI was 2.3% and 3.7%, respectively, for reader A (P = 0.580), and 0.5% and 4.2%, respectively, for reader B (P = 0.033). There were 142 lesions judged as equivocal in at least one of the imaging studies by either reader. Sixty-eight lesions (49%) were recognized as malignant by any of the alternative assessments; among these, 61were confirmed as malignant. Conclusion Gd-EOB-MRI was superior to CE-CT for detecting liver lesions. The false-positive rate was low for both modalities. Equivocal lesions were encountered frequently, but routine use of alternate modalities and evaluation by 2 or more radiologists could enhance the accuracy of diagnosis.