Improved Generalizability in Medical Computer Vision: Hyperbolic Deep Learning in Multi-Modality Neuroimaging

Journal of Imaging, vol. 10, pp. 319

Abstract

Deep learning has shown significant value in automating radiological diagnostics but can be limited by a lack of generalizability to external datasets. Leveraging the geometric principles of non-Euclidean space, certain geometric deep learning approaches may offer an alternative means of improving model generalizability. This study investigates the potential advantages of hyperbolic convolutional neural networks (HCNNs) over traditional convolutional neural networks (CNNs) in neuroimaging tasks. We conducted a comparative analysis of HCNNs and CNNs across various medical imaging modalities and diseases, with a focus on a compiled multi-modality neuroimaging dataset. The models were assessed for their performance parity, robustness to adversarial attacks, semantic organization of embedding spaces, and generalizability. Zero-shot evaluations were also performed with ischemic stroke non-contrast CT images. HCNNs matched CNNs' performance in less complex settings and demonstrated superior semantic organization and robustness to adversarial attacks. While HCNNs equaled CNNs in out-of-sample datasets identifying Alzheimer's disease, in zero-shot evaluations, HCNNs outperformed CNNs and radiologists. HCNNs deliver enhanced robustness and organization in neuroimaging data. This likely underlies why, while HCNNs perform similarly to CNNs with respect to in-sample tasks, they confer improved generalizability. Nevertheless, HCNNs encounter efficiency and performance challenges with larger, complex datasets. These limitations underline the need for further optimization of HCNN architectures. HCNNs present promising improvements in generalizability and resilience for medical imaging applications, particularly in neuroimaging. Despite facing challenges with larger datasets, HCNNs enhance performance under adversarial conditions and offer better semantic organization, suggesting valuable potential in generalizable deep learning models in medical imaging and neuroimaging diagnostics.

Authors 10

1. Cyrus Ayubcha

   Brigham and Women's Hospital · Harvard University

   Affiliation as printed

   Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA

   Harvard Medical School, Boston, MA 02115, USA

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115, USA

2. Sulaiman Sajed

   Boston University · Brigham and Women's Hospital · Harvard University

   Affiliation as printed

   Department of Medical Science, Boston University, Boston, MA 02118, USA

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115, USA

3. Chady Omara Aachen Department of Neurosurgery, Leiden University Medical Center

   Brigham and Women's Hospital · Harvard University · Leiden University Medical Center

   Affiliation as printed

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA

   Department of Neurosurgery, Leiden University Medical Center, 2333 ZA Leiden, The Netherlands

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115, USA

4. Anna Baukje Lebouille-Veldman Aachen Department of Neurosurgery, Leiden University Medical Center

   Brigham and Women's Hospital · Harvard University · Leiden University Medical Center

   Affiliation as printed

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA

   Department of Neurosurgery, Leiden University Medical Center, 2333 ZA Leiden, The Netherlands

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115, USA

5. Shashi B. Singh

   Stanford Medicine · Stanford University

   Affiliation as printed

   Stanford University School of Medicine, Stanford, CA 94305, USA

6. Yashas Ullas Lokesha

   Sri Devaraj Urs Medical College · Sri Devaraj Urs Academy of Higher Education and Research

   Affiliation as printed

   Department of Radiology, Sri Devaraj Urs Medical College, Kolar 563101, India

7. Alex X. Liu

   Temple University

   Affiliation as printed

   Department of Computer and Information Sciences, Temple University, Philadelphia, PA 19122, USA

8. Mohammad Ali Aziz‐Sultan

   Brigham and Women's Hospital · Harvard University

   Affiliation as printed

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115, USA

9. Timothy R. Smith

   Brigham and Women's Hospital · Harvard University

   Affiliation as printed

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA

   Department of Neurosurgery, Cushing Neurosurgical Outcomes Center, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115, USA

10. Andrew L. Beam corresponding

    Harvard University

    Affiliation as printed

    Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA