Scholarlink

Purpose <p>This study evaluated transfer learning in classifying skeletal metastases on bone scintigraphy. The primary objective was to assess its performance in detecting skeletal metastases, while the secondary objective was to compare its performance to human readers.</p> Methods <p>A total of 2,510 patients with known malignancies were included – 2,368 retrospectively recruited and 142 prospectively enrolled. Scans were categorized as normal, benign-degenerative, or metastasis based on clinical consensus, follow-up, biopsy, radiology, or SPECT/CT findings. The retrospective data were randomly divided into training (1,895) and validation (473) sets, while the prospective cohort served as an independent testing set. Google’s InceptionV3 was used for image embedding, and 13 supervised ML algorithms were tested. The Log Loss value of a random classifier was used to select the optimal models for testing, while Stuart-Maxwell test compared models’ performance to human readers.</p> Results <p>Eight ML models with Log Loss value less than that of the random classifier achieved AUCs &gt; 0.900 on training and validation, with all but one (Support Vector Machine) maintaining AUCs &gt; 0.900 on testing. Logistic Regression performed best (≥ 0.993 in all metrics), while Neural Networks, Gradient Boosting, and Random Forest also demonstrated robust performance (≥ 0.817 in all metrics). Notably, ML models interpreted 142 images in 0.027–1.770&#xa0;s compared to 10.07–18.00&#xa0;min by human readers; less experienced readers performed significantly worse (<i>P</i> ≤ 0.002) than the models, whereas experienced reader’s performance was comparable (<i>P</i> ≥ 0.280).</p> Conclusion <p>Transfer learning demonstrates commendable performance in classifying skeletal metastases on bone scintigraphy, outperforming less experienced readers while matching experienced reader’s performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transfer Learning in Bone Scintigraphy: Superior to Early-career Readers and Comparable to Experts in Classifying Skeletal Metastasis

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Vikrant Kumar Goenka,
  • Asem Rangita Chanu,
  • Ritwik Wakankar,
  • Mohammad Umar,
  • Bangkim Chandra Khangembam,
  • Chetan Patel,
  • Rakesh Kumar

摘要

Purpose

This study evaluated transfer learning in classifying skeletal metastases on bone scintigraphy. The primary objective was to assess its performance in detecting skeletal metastases, while the secondary objective was to compare its performance to human readers.

Methods

A total of 2,510 patients with known malignancies were included – 2,368 retrospectively recruited and 142 prospectively enrolled. Scans were categorized as normal, benign-degenerative, or metastasis based on clinical consensus, follow-up, biopsy, radiology, or SPECT/CT findings. The retrospective data were randomly divided into training (1,895) and validation (473) sets, while the prospective cohort served as an independent testing set. Google’s InceptionV3 was used for image embedding, and 13 supervised ML algorithms were tested. The Log Loss value of a random classifier was used to select the optimal models for testing, while Stuart-Maxwell test compared models’ performance to human readers.

Results

Eight ML models with Log Loss value less than that of the random classifier achieved AUCs > 0.900 on training and validation, with all but one (Support Vector Machine) maintaining AUCs > 0.900 on testing. Logistic Regression performed best (≥ 0.993 in all metrics), while Neural Networks, Gradient Boosting, and Random Forest also demonstrated robust performance (≥ 0.817 in all metrics). Notably, ML models interpreted 142 images in 0.027–1.770 s compared to 10.07–18.00 min by human readers; less experienced readers performed significantly worse (P ≤ 0.002) than the models, whereas experienced reader’s performance was comparable (P ≥ 0.280).

Conclusion

Transfer learning demonstrates commendable performance in classifying skeletal metastases on bone scintigraphy, outperforming less experienced readers while matching experienced reader’s performance.