Melola Discovery

← Melola Discovery

Akses terbuka Jurnal SINTA 1

Knowledge Distillation from ResNet-50 to MobileNetV3-Small for Breast Cancer Classification in Ultrasound Images

JURNAL INFOTEL · Vol. 18 No. 3 (2026) · hlm. 472–489

DOI
10.20895/infotel.v18i3.10773
ISSN
2085-3688, 2460-0997
Terbit
2026-08-31
Penerbit
LPPM Institut Teknologi Telkom Purwokerto
Bahasa
en
Lisensi
creativecommons.org/licenses/by-nc-sa/4.0/

Abstrak

Early detection of breast cancer is essential for reducing mortality, but deploying large deep learning models in resource-constrained clinical settings remains challenging. This study evaluates knowledge distillation (KD) from a ResNet-50 teacher to a MobileNetV3-Small student for classifying breast ultrasound images as benign, malignant, or normal, compared with the same student trained without KD. We partitioned the Breast Ultrasound Images Dataset (BUSI), containing 780 images from 600 patients, using a fixed, stratified 70/15/15 image-level split. Unavailable patient identifiers prevented patient-level splitting, constituting a study limitation. We applied Total Variation Chambolle denoising before training. Grid search selected temperature T = 2 and weighting coefficient α = 0.5, achieving 88.89% validation accuracy. Across five random seeds, the KD student achieved 82.05% ±1.60% test accuracy and 80.93% ±2.04% macro F1-score, compared with 82.74% ±1.64% and 81.88% ±2.17% for the no-KD student, and 83.59% ±4.25% and 82.41% ±5.62% for the teacher. Equivalence testing supported KD–teacher equivalence within a ±5-percentage-point margin (p=0.0498). Paired Wilcoxon tests found no evidence of improvement over the no-KD baseline (p=0.625) or denoising-free training (p=0.4375). The student had 15.5 times fewer parameters and approximately 2.5 times faster inference than the teacher. These findings support compact models for breast ultrasound classification but provide no evidence that KD outperforms simple fine-tuning of the same lightweight architecture on this dataset.

Kata kunci

Jaringan kolaborasi para penulis

18 penulis · 36 tautan kepenulisan bersama · 12 lintas organisasi

Arahkan kursor atau ketuk sebuah nama untuk menyorot rekan penulisnya. Klik dua kali untuk membuka halaman penulis.

  • Universitas Ma Chung
  • Sriwijaya University
  • Institut Teknologi Telkom Purwokerto
  • Universitas Kaltara
  • Universitas Gadjah Mada
  • Indonesia University of Education
  • Afiliasi tidak diketahui
  • Organisasi berbeda
  • Organisasi sama
  • Afiliasi tidak diketahui
  • Garis lebih tebal = lebih banyak artikel bersama · lingkaran lebih besar = lebih banyak artikel
Tampilkan sebagai tabel
Rekan penulisOrganisasiArtikel bersama
Hendry SetiawanUniversitas Ma Chung1
Ardian KurniawanUniversitas Ma Chung1
Oesman Hendra KelanaUniversitas Gadjah Mada1
Hermawan WirasantosaUniversitas Ma Chung1
Kestrilia Rega PriliantiUniversitas Ma Chung1
Romy Budhi WidodoAfiliasi tidak diketahui1
Paulus Lucky Tirma IrawanAfiliasi tidak diketahui1
Ratna Dwi ChristyantiAfiliasi tidak diketahui1
Imam AsrowardiAfiliasi tidak diketahui1
Eko SubyantoroAfiliasi tidak diketahui1
Septafiansyah Dwi PutraAfiliasi tidak diketahui1
Ari WidodoIndonesia University of Education1
Ahmad Fali OklilasSriwijaya University1
Milda KamiliaSriwijaya University1
Abdurahman AbdurahmanSriwijaya University1

Afiliasi berasal dari OpenAlex; metadata OAI-PMH jurnal tidak memuatnya. Penulis tanpa ID OpenAlex dicocokkan berdasarkan nama, sehingga dua orang bisa tergabung atau satu orang bisa terpisah.

Sumber metadata: OpenAlex API (jurnal SINTA, ADR 0048) https://openalex.org/S4210188896, pertama di-harvest 19 Nov 2026, terakhir terlihat 19 Nov 2026.
Metadata asli (oai_dc)
<oai_dc:dc xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:oai_dc="http://www.openarchives.org/OAI/2.0/oai_dc/"><dc:title xml:lang="en">Knowledge Distillation from ResNet-50 to MobileNetV3-Small for Breast Cancer Classification in Ultrasound Images</dc:title><dc:creator>Windra Swastika</dc:creator><dc:creator>Bita Parga Zen</dc:creator><dc:creator>Ratna Dwi Christyanti</dc:creator><dc:subject>knowledge distillation</dc:subject><dc:subject>breast cancer classification</dc:subject><dc:subject>breast ultrasound images</dc:subject><dc:subject>MobileNetV3-Small</dc:subject><dc:subject>ResNet50</dc:subject><dc:subject>lightweight models</dc:subject><dc:subject>model compression</dc:subject><dc:subject>BUSI dataset</dc:subject><dc:subject>total variation denoising</dc:subject><dc:subject>temperature scaling</dc:subject><dc:subject>benign versus malignant differentiation</dc:subject><dc:description xml:lang="en">Early detection of breast cancer is essential for reducing mortality, but deploying large deep learning models in resource-constrained clinical settings remains challenging. This study evaluates knowledge distillation (KD) from a ResNet-50 teacher to a MobileNetV3-Small student for classifying breast ultrasound images as benign, malignant, or normal, compared with the same student trained without KD. We partitioned the Breast Ultrasound Images Dataset (BUSI), containing 780 images from 600 patients, using a fixed, stratified 70/15/15 image-level split. Unavailable patient identifiers prevented patient-level splitting, constituting a study limitation. We applied Total Variation Chambolle denoising before training. Grid search selected temperature T = 2 and weighting coefficient α = 0.5, achieving 88.89% validation accuracy. Across five random seeds, the KD student achieved 82.05% ±1.60% test accuracy and 80.93% ±2.04% macro F1-score, compared with 82.74% ±1.64% and 81.88% ±2.17% for the no-KD student, and 83.59% ±4.25% and 82.41% ±5.62% for the teacher. Equivalence testing supported KD–teacher equivalence within a ±5-percentage-point margin (p=0.0498). Paired Wilcoxon tests found no evidence of improvement over the no-KD baseline (p=0.625) or denoising-free training (p=0.4375). The student had 15.5 times fewer parameters and approximately 2.5 times faster inference than the teacher. These findings support compact models for breast ultrasound classification but provide no evidence that KD outperforms simple fine-tuning of the same lightweight architecture on this dataset.</dc:description><dc:publisher>LPPM Institut Teknologi Telkom Purwokerto</dc:publisher><dc:date>2026-08-31</dc:date><dc:type>info:eu-repo/semantics/article</dc:type><dc:identifier>https://doi.org/10.20895/infotel.v18i3.10773</dc:identifier><dc:identifier>10.20895/infotel.v18i3.10773</dc:identifier><dc:source>JURNAL INFOTEL; Vol 18 No 3 (2026); 472-489</dc:source><dc:source>2085-3688</dc:source><dc:source>2460-0997</dc:source><dc:language>en</dc:language><dc:rights>https://creativecommons.org/licenses/by-nc-sa/4.0/</dc:rights></oai_dc:dc>