跳转至

文章背景与核心概要

在低收入和中等收入国家（LMIC），医疗服务的提供往往面临重大阻碍，这主要是由于 MRI 和超声等先进影像设备的频繁停机，且当地严重缺乏专业的生物医学工程技术支持。为了解决这一挑战，研究人员开发了一个多模态医疗设备维护问答（QA）框架，并针对专门的技术故障排除任务对医疗基础模型进行了微调。

该团队在九个 LMIC 进行了跨国调查的指导下，整理了 MRI 和超声系统的技术手册，构建了包含 10,294 个高质量过滤问答上下文对的 INGENZI_DatasetV1 数据集。利用基于 QLoRA 的参数高效微调方法，他们调整了 MedGemma-4b-it 模型，使其能够解释系统错误日志并生成分步设备修理指令。实验结果表明，微调后的模型在多项评估指标上取得了显著提升，为资源受限环境下的 AI 辅助诊断与维护工具奠定了可靠的基础。

From Manuals to Maintenance: Fine-Tuning MedGemma for Multi-Modal Imaging System Support in Low-Resource Settings

Executive Summary

Healthcare delivery in low- and middle-income countries (LMICs) often faces critical barriers due to the frequent downtime of advanced imaging devices (such as MRI and ultrasound systems). This issue is compounded by a severe shortage of specialized biomedical engineering support.

To address this challenge, researchers developed a multi-modality medical equipment maintenance question-answering (QA) framework. Guided by a multi-country survey across nine LMICs, the team curated technical manuals to build the INGENZI_DatasetV1 (comprising 10,294 filtered QA-context pairs). Using QLoRA-based parameter-efficient fine-tuning, they adapted the MedGemma-4b-it model to accurately interpret system error logs and generate step-by-step equipment repair instructions. The resulting fine-tuned model demonstrated significant improvements over the baseline across multiple evaluation metrics, providing a scalable, reliable foundation for AI-assisted diagnostic and maintenance tools in resource-constrained environments.

Healthcare delivery in low- and middle-income countries (LMICs) often faces critical barriers due to the frequent downtime of advanced imaging devices (such as MRI and ultrasound systems). This issue is compounded by a severe shortage of specialized biomedical engineering support.

To address this challenge, researchers developed a multi-modality medical equipment maintenance question-answering (QA) framework. Guided by a multi-country survey across nine LMICs, the team curated technical manuals to build the INGENZI_DatasetV1 (comprising 10,294 filtered QA-context pairs). Using QLoRA-based parameter-efficient fine-tuning, they adapted the MedGemma-4b-it model to accurately interpret system error logs and generate step-by-step equipment repair instructions. The resulting fine-tuned model demonstrated significant improvements over the baseline across multiple evaluation metrics, providing a scalable, reliable foundation for AI-assisted diagnostic and maintenance tools in resource-constrained environments.

Article Metadata

  • arXiv ID: arXiv:2608.08896 [cs.AI]
  • Submitted On: August 9, 2026
  • Primary Subject: Artificial Intelligence (cs.AI)
  • Conference Status: Accepted at the AFRICAI 2026 Workshop (a satellite event at MICCAI 2026); to appear in Springer Lecture Notes in Computer Science (LNCS).
  • DOI: 10.48550/arXiv.2608.08896
  • arXiv ID: arXiv:2608.08896 [cs.AI]
  • Submitted On: August 9, 2026
  • Primary Subject: Artificial Intelligence (cs.AI)
  • Conference Status: Accepted at the AFRICAI 2026 Workshop (a satellite event at MICCAI 2026); to appear in Springer Lecture Notes in Computer Science (LNCS).
  • DOI: 10.48550/arXiv.2608.08896

Authors

  • Bernes Lorier Atabonfack
  • Zion Kongbi Nfo
  • Ahmed Tahiru Issah
  • Tolulope Olusuyi
  • Clemence Ingabire
  • Mohammed Hardi Abdul Baaki
  • Mawuli Deku
  • Abdulrazaq Zubair
  • Alyasaa Anas
  • Raymond Confidence
  • Maruf Adewole
  • Udunna C. Anazodo
  • Bernes Lorier Atabonfack
  • Zion Kongbi Nfo
  • Ahmed Tahiru Issah
  • Tolulope Olusuyi
  • Clemence Ingabire
  • Mohammed Hardi Abdul Baaki
  • Mawuli Deku
  • Abdulrazaq Zubair
  • Alyasaa Anas
  • Raymond Confidence
  • Maruf Adewole
  • Udunna C. Anazodo

Abstract

影像设备停机是低收入和中等收入国家（LMIC）医疗服务提供的主要障碍，这通常是由专业生物医学工程支持渠道有限所导致的。我们提出了一个多模态医疗设备维护问答（QA）框架，并展示了针对专门技术故障排除任务对医疗基础模型的微调。

在九个 LMIC 跨国调查的指导下，我们整理了 MRI 和超声系统的技术手册，生成了 INGENZI_DatasetV1，其中包含 10,294 个高质量、经过过滤的问答上下文对。利用基于 QLoRA 的参数高效微调，我们调整了 MedGemma-4b-it 模型，使其能够解释系统错误日志并生成分步设备修理指令。

与基线模型相比，微调后的系统在各项指标上都取得了实质性提升，包括： * F1 Score: \(0.22 \rightarrow 0.38\) * ROUGE-2: \(0.18 \rightarrow 0.41\) * BERTScore F1: \(0.86 \rightarrow 0.91\)

这些指标的提升表明，该模型能够针对新的故障排除查询，生成明显更精确且程序正确的技术回应。这项工作为资源受限环境下的 AI 辅助诊断和维护工具奠定了可靠的基础。

Imaging device downtime is a major barrier to healthcare delivery in low- and middle-income countries (LMICs), often driven by limited access to specialized biomedical engineering support. We present a multi-modality medical equipment maintenance question-answering (QA) framework and demonstrate the fine-tuning of a medical foundation model for specialized technical troubleshooting tasks.

Guided by a multi-country survey across nine LMICs, we curated technical manuals from MRI and ultrasound systems to generate the INGENZI_DatasetV1, containing 10,294 high-quality, filtered QA-context pairs. Using QLoRA-based parameter-efficient fine-tuning, we adapted the MedGemma-4b-it model to interpret system error logs and generate step-by-step equipment repair instructions.

Compared to the baseline model, the fine-tuned system achieved substantial improvements across metrics, including: * F1 Score: \(0.22 \rightarrow 0.38\) * ROUGE-2: \(0.18 \rightarrow 0.41\) * BERTScore F1: \(0.86 \rightarrow 0.91\)

These metric gains demonstrate that the model generates significantly more precise and procedurally accurate technical responses to new troubleshooting queries. This work establishes a reliable foundation for AI-assisted diagnostic and maintenance tools in resource-constrained settings.

Full-Text & Resource Links