SQCDL-WIoT: a scalable quantum convolutional deep learning framework with wearable IoT integration for breast cancer screening

Abstract Deep learning techniques have greatly improved the ability to perform automatic diagnosis of breast cancer through use of computers; however, existing techniques are generally complex and costly, do not well incorporate multiple data modalities, and cannot be deployed clinically in real time. This paper proposes SQCDL-WIoT, which is a scalable, wearable-based internet of things (IoT)-enabled quantum convolutional deep learning (C-DL) framework capable of multi-class breast cancer screening. The proposed method includes two VQCs with five qubits each that can be embedded into a ResNet-50 back bone. In addition to being able to process images from histopathology and wearable-based bio sensor signals as input, the method also enables the incorporation of both wearable based bio sensor signal processing (i.e., wearable-IoT) and clinical variables via the Squeeze-and-Excitation (SE) attention mechanism. Thus, this proposed method is capable of hybrid quantum-classical training. Using stratified k-folds cross-validation (where k = 5) on the BreaKHis dataset and zero-shot transfer learning on the CBIS-DDSM dataset, SQCDL-WIoT demonstrated the highest accuracy (98.14%) with the highest sensitivity (macro-sensitivity = 97.63%), specificity (macro-specificity = 98.87%), and area under curve (AUC = 0.9891) when compared to all other baseline methods at 40x magnification on BreaKHis. Additionally, SQCDL-WIoT demonstrated 93.41% accuracy on the CBIS-DDSM dataset via zero shot learning, indicating good cross-domain generalization. Finally, the wearable IoT sub-system demonstrated a mean real-time edge-inference latency of 378 ± 12 milliseconds on an ARM Cortex-A72 processor, meeting the defined clinical threshold for response of < 400 milliseconds. The results of ablation studies indicated that both quantum convolution and SE attention contributed to complementary and synergetic increases in performance compared to classical ResNet-50 baseline models. Overall, the collective results presented here indicate that the proposed hybrid quantum-classical DL framework combined with wearable-based IoT sensing may represent a viable solution to scalable and intelligent breast cancer screening.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-13
DOI
https://doi.org/10.1038/s41598-026-73447-1
Primary Topic
AI in cancer detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

SQCDL-WIoT: a scalable quantum convolutional deep learning framework with wearable IoT integration for breast cancer screening

Anju Susan George, Kama Ramudu, Swapna Saturi, Nainavarapu Radha et al.
3 links
Scientific Reports
AI in cancer detection
article

SQCDL-WIoT: a scalable quantum convolutional deep learning framework with wearable IoT integration for breast cancer screening

Anju Susan George, Kama Ramudu, Swapna Saturi, Nainavarapu Radha, H. K. Prasad Katamreddi, Stuti Pandey, Boddu Ramadevi
article en

Abstract

Abstract Deep learning techniques have greatly improved the ability to perform automatic diagnosis of breast cancer through use of computers; however, existing techniques are generally complex and costly, do not well incorporate multiple data modalities, and cannot be deployed clinically in real time. This paper proposes SQCDL-WIoT, which is a scalable, wearable-based internet of things (IoT)-enabled quantum convolutional deep learning (C-DL) framework capable of multi-class breast cancer screening. The proposed method includes two VQCs with five qubits each that can be embedded into a ResNet-50 back bone. In addition to being able to process images from histopathology and wearable-based bio sensor signals as input, the method also enables the incorporation of both wearable based bio sensor signal processing (i.e., wearable-IoT) and clinical variables via the Squeeze-and-Excitation (SE) attention mechanism. Thus, this proposed method is capable of hybrid quantum-classical training. Using stratified k-folds cross-validation (where k = 5) on the BreaKHis dataset and zero-shot transfer learning on the CBIS-DDSM dataset, SQCDL-WIoT demonstrated the highest accuracy (98.14%) with the highest sensitivity (macro-sensitivity = 97.63%), specificity (macro-specificity = 98.87%), and area under curve (AUC = 0.9891) when compared to all other baseline methods at 40x magnification on BreaKHis. Additionally, SQCDL-WIoT demonstrated 93.41% accuracy on the CBIS-DDSM dataset via zero shot learning, indicating good cross-domain generalization. Finally, the wearable IoT sub-system demonstrated a mean real-time edge-inference latency of 378 ± 12 milliseconds on an ARM Cortex-A72 processor, meeting the defined clinical threshold for response of < 400 milliseconds. The results of ablation studies indicated that both quantum convolution and SE attention contributed to complementary and synergetic increases in performance compared to classical ResNet-50 baseline models. Overall, the collective results presented here indicate that the proposed hybrid quantum-classical DL framework combined with wearable-based IoT sensing may represent a viable solution to scalable and intelligent breast cancer screening.

Scientific Reports
Kakatiya University (IN), Vaagdevi College of Engineering (IN), Aditya University (IN), Manipal University Jaipur (IN)
Good health and well-being
AI in cancer detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon