Artificial Intelligence with Statistical Confidence Scores for Detection of Acute or Subacute Hemorrhage on Noncontrast CT Head Scans.

Purpose: To present a method that automatically detects, subtypes, and locates acute or subacute intracranial hemorrhage (ICH) on noncontrast CT (NCCT) head scans; generates detection confidence scores to identify high-confidence data subsets with higher accuracy; and improves radiology worklist prioritization. Such scores may enable clinicians to better use artificial intelligence (AI) tools. Materials and Methods: This retrospective study included 46 057 studies from seven "internal" centers for development (training, architecture selection, hyperparameter tuning, and operating-point calibration; n = 25 946) and evaluation (n = 2947) and three "external" centers for calibration (n = 400) and evaluation (n = 16 764). Internal centers contributed developmental data, whereas external centers did not. Deep neural networks predicted the presence of ICH and subtypes (intraparenchymal, intraventricular, subarachnoid, subdural, and/or epidural hemorrhage) and segmentations per case. Two ICH confidence scores are discussed: a calibrated classifier entropy score and a Dempster-Shafer score. Evaluation was completed by using receiver operating characteristic curve analysis and report turnaround time (RTAT) modeling on the evaluation set and on confidence score-defined subsets using bootstrapping. Results: The areas under the receiver operating characteristic curve for ICH were 0.97 (0.97, 0.98) and 0.95 (0.94, 0.95) on internal and external center data, respectively. On 80% of the data stratified by calibrated classifier and Dempster-Shafer scores, the system improved the Youden indexes, increasing them from 0.84 to 0.93 (calibrated classifier) and from 0.84 to 0.92 (Dempster-Shafer) for internal centers and increasing them from 0.78 to 0.88 (calibrated classifier) and from 0.78 to 0.89 (Dempster-Shafer) for external centers (P < .001). Models estimated shorter RTAT for AI-prioritized worklists with confidence measures than for AI-prioritized worklists without confidence measures, shortening RTAT by 27% (calibrated classifier) and 27% (Dempster-Shafer) for internal centers and shortening RTAT by 25% (calibrated classifier) and 27% (Dempster-Shafer) for external centers (P < .001). Conclusion: AI that provided statistical confidence measures for ICH detection on NCCT scans reliably detected and subtyped hemorrhages, identified high-confidence predictions, and improved worklist prioritization in simulation.Keywords: CT, Head/Neck, Hemorrhage, Convolutional Neural Network (CNN) Supplemental material is available for this article. © RSNA, 2022.

Spatial heterogeneity of oxygenation and haemodynamics in breast cancer resolved in vivo by conical multispectral optoacoustic mesoscopy.

The characteristics of tumour development and metastasis relate not only to genomic heterogeneity but also to spatial heterogeneity, associated with variations in the intratumoural arrangement of cell populations, vascular morphology and oxygen and nutrient supply. While optical (photonic) microscopy is commonly employed to visualize the tumour microenvironment, it assesses only a few hundred cubic microns of tissue. Therefore, it is not suitable for investigating biological processes at the level of the entire tumour, which can be at least four orders of magnitude larger. In this study, we aimed to extend optical visualization and resolve spatial heterogeneity throughout the entire tumour volume. We developed an optoacoustic (photoacoustic) mesoscope adapted to solid tumour imaging and, in a pilot study, offer the first insights into cancer optical contrast heterogeneity in vivo at an unprecedented resolution of <50 µm throughout the entire tumour mass. Using spectral methods, we resolve unknown patterns of oxygenation, vasculature and perfusion in three types of breast cancer and showcase different levels of structural and functional organization. To our knowledge, these results are the most detailed insights of optical signatures reported throughout entire tumours in vivo, and they position optoacoustic mesoscopy as a unique investigational tool linking microscopic and macroscopic observations.

High-Resolution Multispectral Optoacoustic Tomography of the Vascularization and Constitutive Hypoxemia of Cancerous Tumors.

Diversity of the design and alignment of illumination and ultrasonic transducers empower the fine scalability and versatility of optoacoustic imaging. In this study, we implement an innovative high-resolution optoacoustic mesoscopy for imaging the vasculature and tissue oxygenation within subcutaneous and orthotopic cancerous implants of mice in vivo through acquisition of tomographic projections over 180° at a central frequency of 24 MHz. High-resolution volumetric imaging was combined with multispectral functional measurements to resolve the exquisite inner structure and vascularization of the entire tumor mass using endogenous and exogenous optoacoustic contrast. Evidence is presented for constitutive hypoxemia within the carcinogenic tissue through analysis of the hemoglobin absorption spectra and distribution. Morphometric readouts obtained with optoacoustic mesoscopy have been verified with high-resolution ultramicroscopic studies. The findings described herein greatly extend the applications of optoacoustic mesoscopy toward structural and multispectral functional measurements of the vascularization and hemodynamics within solid tumors in vivo and are of major relevance to basic and preclinical oncological studies in small animal models.

Optical mesoscopy without the scatter: broadband multispectral optoacoustic mesoscopy.

Optical mesoscopy extends the capabilities of biological visualization beyond the limited penetration depth achieved by microscopy. However, imaging of opaque organisms or tissues larger than a few hundred micrometers requires invasive tissue sectioning or chemical treatment of the specimen for clearing photon scattering, an invasive process that is regardless limited with depth. We developed previously unreported broadband optoacoustic mesoscopy as a tomographic modality to enable imaging of optical contrast through several millimeters of tissue, without the need for chemical treatment of tissues. We show that the unique combination of three-dimensional projections over a broad 500 kHz-40 MHz frequency range combined with multi-wavelength illumination is necessary to render broadband multispectral optoacoustic mesoscopy (2B-MSOM) superior to previous optical or optoacoustic mesoscopy implementations.

Ultra-wideband three-dimensional optoacoustic tomography.

Broadband optoacoustic waves generated by biological tissues excited with nanosecond laser pulses carry information corresponding to a wide range of geometrical scales. Typically, the frequency content present in the signals generated during optoacoustic imaging is much larger compared to the frequency band captured by common ultrasonic detectors, the latter typically acting as bandpass filters. To image optical absorption within structures ranging from entire organs to microvasculature in three dimensions, we implemented optoacoustic tomography with two ultrasound linear arrays featuring a center frequency of 6 and 24 MHz, respectively. In the present work, we show that complementary information on anatomical features could be retrieved and provide a better understanding on the localization of structures in the general anatomy by analyzing multi-bandwidth datasets acquired on a freshly excised kidney.

High-resolution optoacoustic mesoscopy with a 24 MHz multidetector translate-rotate scanner.

Optoacoustic (photoacoustic) mesoscopy aims at high-resolution optical imaging of anatomical, functional, and cellular parameters at depths that go well beyond those of optical-resolution optical or optoacoustic microscopy i.e., reaching several millimeters in depth. The approach utilizes tomography to achieve ultrasonic-diffraction resolution and operates at high-ultrasound frequencies (20 to 200 MHz) induced by few-nanosecond laser pulse excitation of tissues. We investigated here the performance of optoacoustic mesoscopy implemented at 24 MHz center frequency and its ability to resolve optical absorption contrast in the mouse kidney ex vivo. The developed system achieved better than 30 µm in-plane resolution and 110 µm elevation resolution over a cylindrical volume of 9-mm diameter and 9-mm height. This unprecedented combination of resolution and depth was achieved by implementing a translate-rotate detection geometry and by tomographic reconstruction. The approach yielded images of optically absorbing structures with a level of detail never-before visualized in an intact mouse kidney and allows insights into their unperturbed architecture. We discuss the ability to offer multispectral acquisitions and enable in vivo imaging.