Saad Abbasi

Mr. Saad Abbasi


Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 September 2019 in Scientific Reports
Verified
Journal Article
Open access
PDF file
SASaad Abbasi
Saad Abbasi
MLMartin Le
Martin Le
BSBazil Sonier
Bazil Sonier
DDDeepak Dinakaran
Deepak Dinakaran
GBGilbert Bigras
Gilbert Bigras
PRParsin Haji Reza
Parsin Haji Reza
Surgical oncologists depend heavily on visual field acuity during cancer resection surgeries for in-situ margin assessment. Clinicians must wait up to two weeks for results from a pathology lab to confirm a post-operative diagnosis, potentially resulting in subsequent treatments. Currently, there are no clinical tools that can visualize diagnostically pertinent tissue information in-situ. Here, we present the first microscopy capable of non-contact label-free visualization of human cellular morphology in a reflection-mode apparatus. This is possible with the recently reported imaging modality called photoacoustic remote sensing microscopy which enables non-contact detection of optical absorption contrast. By taking advantage of the 266-nanometer optical absorption peak of DNA, photoacoustic remote sensing is efficacious in recovering qualitatively similar nuclear information in comparison to that provided by the hematoxylin stain in the gold-standard hematoxylin and eosin (H&E) prepared samples. A photoacoustic remote sensing system was employed utilizing a 266-nanometer pulsed excitation beam to induce photoacoustic pressures within the sample resulting in refractive index modulation of the optical absorber. A 1310-nanometer continuous-wave interrogation beam detects these perturbed regions as back reflected intensity variations due to the changes in the local optical properties. Using this technique, clinically useful histologic images of human tissue samples including breast cancer (invasive ductal carcinoma), tonsil, gastrointestinal, and pancreatic tissue images were formed. These were qualitatively comparable to standard H&E prepared samples.

14 July 2021 in Sensors
Verified
Journal Article
Open access
PDF file
SASaad Abbasi
Saad Abbasi
MFMahmoud Famouri
Mahmoud Famouri
MSMohammad Javad Shafiee
Mohammad Javad Shafiee
Alexander WongAlexander Wong
Alexander Wong
Human operators often diagnose industrial machinery via anomalous sounds. Given the new advances in the field of machine learning, automated acoustic anomaly detection can lead to reliable maintenance of machinery. However, deep learning-driven anomaly detection methods often require an extensive amount of computational resources prohibiting their deployment in factories. Here we explore a machine-driven design exploration strategy to create OutlierNets, a family of highly compact deep convolutional autoencoder network architectures featuring as few as 686 parameters, model sizes as small as 2.7 KB, and as low as 2.8 million FLOPs, with a detection accuracy matching or exceeding published architectures with as many as 4 million parameters. The architectures are deployed on an Intel Core i5 as well as a ARM Cortex A72 to assess performance on hardware that is likely to be used in industry. Experimental results on the model’s latency show that the OutlierNet architectures can achieve as much as 30× lower latency than published networks.

1 October 2019 in Biomedical Optics Express
Verified
Journal Article
Closed access
PDF file
SASaad Abbasi
Saad Abbasi
MLMartin Le
Martin Le
BSBazil Sonier
Bazil Sonier
KBKevan Bell
Kevan Bell
DDDeepak Dinakaran
Deepak Dinakaran
PRParsin Haji Reza
Parsin Haji Reza
Identifying positive surgical margins after resection of cancer often triggers re-excision and adjuvant treatments. Incomplete initial resections result in poorer patient outcomes, psychological and financial stress to the patient and increased healthcare costs. Surgical margins are typically assessed post-operatively using time consuming and expensive slide-based histopathology tissue analysis. Currently, a real-time non-contact virtual histology-like intraoperative margin assessment tool is not available. To address this need, we have developed a non-contact multi-wavelength reflection-mode, photoacoustic remote sensing (PARS) microscope demonstrating chromophore selective contrast in human tissues. We show the capabilities of multi-wavelength PARS microscopy utilizing both 266 nm and 532 nm excitation wavelengths and a 1310 nm detection wavelength. Cell nuclei and hemoglobin were visualized at the cellular scale without the addition of exogenous contrast agents. These works provide a critical step towards a virtual histology tool to provide intraoperative histology-like information in living tissue.

14 February 2020 in Sensors
Verified
Journal Article
Open access
PDF file
SASaad Abbasi
Saad Abbasi
KBKevan Bell
Kevan Bell
PRParsin Haji Reza
Parsin Haji Reza
Mechanical stages are routinely used to scan large expanses of biological specimens in photoacoustic imaging. This is primarily due to the limited field of view (FOV) provided by optical scanning. However, stage scanning becomes impractical at higher scanning speeds, or potentially unfeasible with heavier samples. Also, the slow scan-rate of the stages makes high resolution scanning a time-consuming process. Some clinical applications such as microsurgery require submicron resolution in a reflection-mode configuration necessitating a method that can acquire large field of views with a small raster scanning step size. In this study, we describe a method that combines mechanical stages with optical scanning for the rapid acquisition of high-resolution large FOVs. Optical scanning is used to acquire small frames in a two-dimensional grid formed by the mechanical stages. These frames are captured with specific overlap for effective image registration. Using a step size of 200 nm, we demonstrate mosaics of carbon fiber networks with FOVs of 0.8 × 0.8 mm2 captured in under 70 s with 1.2 µm image resolution. Larger mosaics yielding an imaging area of 3 × 3 mm2 are also shown. The method is validated by imaging a 1 × 1 mm2 section of unstained histopathological human tissue.

17 September 2020 in arXiv
Verified
Preprint
Open access
PDF file
KBKevan Bell
Kevan Bell
SASaad Abbasi
Saad Abbasi
DDDeepak Dinakaran
Deepak Dinakaran
MTMuba Taher
Muba Taher
GBGilbert Bigras
Gilbert Bigras
PRParsin Haji Reza
Parsin Haji Reza
Histological visualizations are critical to clinical disease management and are fundamental to biological understanding. However, current approaches that rely on bright-field microscopy require extensive tissue preparation prior to imaging. These processes are labor intensive and contribute to delays in clinical feedback that can extend to two to three weeks for standard paraffin-embedded tissue preparation and interpretation. Here, we present a label-free reflection-mode imaging modality that reveals cellular-scale morphology by detecting intrinsic endogenous contrast. We accomplish this with the novel photoacoustic remote sensing (PARS) detection system that permits non-contact optical absorption contrast to be extracted from thick and opaque biological targets with optical resolution. PARS was examined both as a rapid assessment tool that is capable of managing large samples (>1 cm2) in under 10 minutes, and as a high contrast imaging modality capable of extracting specific biological contrast to simulate conventional histological stains such as hematoxylin and eosin (H&E). The capabilities of the proposed method are demonstrated in a variety of human tissue preparations including formalin-fixed paraffin-embedded tissue blocks and unstained slides sectioned from these blocks, including normal and neoplastic human brain, and breast epithelium involved with breast cancer. Similarly, PARS images of human skin prepared by frozen section clearly demonstrated basal cell carcinoma and normal human skin tissue. Finally, we imaged unprocessed murine kidney and achieved histologically relevant subcellular morphology in fresh tissue. This represents a vital step towards an effective real-time clinical microscope that overcomes the limitations of standard histopathologic tissue preparations and enables real-time pathology assessment.

21 April 2023 in arXiv
Verified
Preprint
Open access
PDF file
Alexander WongAlexander Wong
Alexander Wong
YWYifan Wu
Yifan Wu
SASaad Abbasi
Saad Abbasi
SNSaeejith Nair
Saeejith Nair
YCYuhao Chen
Yuhao Chen
MSMohammad Javad Shafiee
Mohammad Javad Shafiee
Multi-task learning has shown considerable promise for improving the performance of deep learning-driven vision systems for the purpose of robotic grasping. However, high architectural and computational complexity can result in poor suitability for deployment on embedded devices that are typically leveraged in robotic arms for real-world manufacturing and warehouse environments. As such, the design of highly efficient multi-task deep neural network architectures tailored for computer vision tasks for robotic grasping on the edge is highly desired for widespread adoption in manufacturing environments. Motivated by this, we propose Fast GraspNeXt, a fast self-attention neural network architecture tailored for embedded multi-task learning in computer vision tasks for robotic grasping. To build Fast GraspNeXt, we leverage a generative network architecture search strategy with a set of architectural constraints customized to achieve a strong balance between multi-task learning performance and embedded inference efficiency. Experimental results on the MetaGraspNet benchmark dataset show that the Fast GraspNeXt network design achieves the highest performance (average precision (AP), accuracy, and mean squared error (MSE)) across multiple computer vision tasks when compared to other efficient multi-task network architecture designs, while having only 17.8M parameters (about >5x smaller), 259 GFLOPs (as much as >5x lower) and as much as >3.15x faster on a NVIDIA Jetson TX2 embedded processor.